WindowServer: Fix xbanan access check

Instead of calculating bit-by-bit crc32, we now calculate a lookup table
during compile time. The old crc32 calculation was taking almost 50% of
the decompression time.

Also handle multiple symbols at once without outputting to user. It is
much more efficient to output many bytes instead of the up to 258 that a
single symbol can decode to :^)

.gitignore

@@ -3,3 +3,4 @@
 build/
 base/
 script/fakeroot-context
+tools/update-image-perms

BAN/include/BAN/Array.h

@@ -35,7 +35,7 @@ namespace BAN
 		constexpr T& front();
 
 		Span<T> span() { return Span(m_data, size()); }
-		const Span<T> span() const { return Span(m_data, size()); }
+		Span<const T> span() const { return Span(m_data, size()); }
 
 		constexpr size_type size() const;
 

BAN/include/BAN/Debug.h

@@ -9,29 +9,35 @@
 #include <BAN/Formatter.h>
 #include <stdio.h>
 
-#define __debug_putchar [](int c) { putc(c, stddbg); }
+#define __debug_putchar [](int c) { putc_unlocked(c, stddbg); }
 
 #define dprintln(...)											\
 	do {														\
+		flockfile(stddbg);										\
 		BAN::Formatter::print(__debug_putchar, __VA_ARGS__);	\
 		BAN::Formatter::print(__debug_putchar,"\n");			\
 		fflush(stddbg);											\
+		funlockfile(stddbg);									\
 	} while (false)
 
 #define dwarnln(...)											\
 	do {														\
+		flockfile(stddbg);										\
 		BAN::Formatter::print(__debug_putchar, "\e[33m");		\
 		BAN::Formatter::print(__debug_putchar, __VA_ARGS__);	\
 		BAN::Formatter::print(__debug_putchar, "\e[m\n");		\
 		fflush(stddbg);											\
+		funlockfile(stddbg);									\
 	} while(false)
 
 #define derrorln(...)											\
 	do {														\
+		flockfile(stddbg);										\
 		BAN::Formatter::print(__debug_putchar, "\e[31m");		\
 		BAN::Formatter::print(__debug_putchar, __VA_ARGS__);	\
 		BAN::Formatter::print(__debug_putchar, "\e[m\n");		\
 		fflush(stddbg);											\
+		funlockfile(stddbg);									\
 	} while(false)
 
 #define dprintln_if(cond, ...)		\

BAN/include/BAN/Heap.h

@@ -0,0 +1,89 @@
+#pragma once
+
+#include <BAN/Iterators.h>
+#include <BAN/Swap.h>
+#include <BAN/Traits.h>
+
+namespace BAN
+{
+
+	namespace detail
+	{
+
+		template<typename It, typename Comp>
+		void heapify_up(It begin, size_t index, Comp comp)
+		{
+			size_t parent = (index - 1) / 2;
+			while (parent < index)
+			{
+				if (comp(*(begin + index), *(begin + parent)))
+					break;
+				swap(*(begin + parent), *(begin + index));
+				index = parent;
+				parent = (index - 1) / 2;
+			}
+		}
+
+		template<typename It, typename Comp>
+		void heapify_down(It begin, size_t index, size_t len, Comp comp)
+		{
+			for (;;)
+			{
+				const size_t lchild = 2 * index + 1;
+				const size_t rchild = 2 * index + 2;
+
+				size_t child = 0;
+				if (lchild < len && !comp(*(begin + lchild), *(begin + index)))
+				{
+					if (rchild < len && !comp(*(begin + rchild), *(begin + lchild)))
+						child = rchild;
+					else
+						child = lchild;
+				}
+				else if (rchild < len && !comp(*(begin + rchild), *(begin + index)))
+					child = rchild;
+				else
+					break;
+
+				swap(*(begin + child), *(begin + index));
+				index = child;
+			}
+		}
+
+	}
+
+	template<typename It, typename Comp = less<it_value_type_t<It>>>
+	void make_heap(It begin, It end, Comp comp = {})
+	{
+		const size_t len = distance(begin, end);
+		if (len <= 1)
+			return;
+
+		size_t index = (len - 2) / 2;
+		while (index < len)
+			detail::heapify_down(begin, index--, len, comp);
+	}
+
+	template<typename It, typename Comp = less<it_value_type_t<It>>>
+	void push_heap(It begin, It end, Comp comp = {})
+	{
+		const size_t len = distance(begin, end);
+		detail::heapify_up(begin, len - 1, comp);
+	}
+
+	template<typename It, typename Comp = less<it_value_type_t<It>>>
+	void pop_heap(It begin, It end, Comp comp = {})
+	{
+		const size_t len = distance(begin, end);
+		swap(*begin, *(begin + len - 1));
+		detail::heapify_down(begin, 0, len - 1, comp);
+	}
+
+	template<typename It, typename Comp = less<it_value_type_t<It>>>
+	void sort_heap(It begin, It end, Comp comp = {})
+	{
+		while (begin != end)
+			pop_heap(begin, end--, comp);
+	}
+
+}

BAN/include/BAN/IPv4.h

@@ -9,6 +9,10 @@ namespace BAN
 
 	struct IPv4Address
 	{
+		constexpr IPv4Address()
+			: IPv4Address(0)
+		{ }
+
 		constexpr IPv4Address(uint32_t u32_address)
 		{
 			raw = u32_address;

BAN/include/BAN/MacroUtils.h

@@ -0,0 +1,46 @@
+#pragma once
+
+#define _ban_count_args_impl(_0, _1, _2, _3, _4, _5, _6, _7, _8, _9, ...) _9
+#define _ban_count_args(...) _ban_count_args_impl(__VA_ARGS__ __VA_OPT__(,) 9, 8, 7, 6, 5, 4, 3, 2, 1, 0)
+
+#define _ban_concat_impl(a, b) a##b
+#define _ban_concat(a, b) _ban_concat_impl(a, b)
+
+#define _ban_stringify_impl(x) #x
+#define _ban_stringify(x) _ban_stringify_impl(x)
+
+#define _ban_fe_0(f)
+#define _ban_fe_1(f, _0)                                 f(0, _0)
+#define _ban_fe_2(f, _0, _1)                             f(0, _0) f(1, _1)
+#define _ban_fe_3(f, _0, _1, _2)                         f(0, _0) f(1, _1) f(2, _2)
+#define _ban_fe_4(f, _0, _1, _2, _3)                     f(0, _0) f(1, _1) f(2, _2) f(3, _3)
+#define _ban_fe_5(f, _0, _1, _2, _3, _4)                 f(0, _0) f(1, _1) f(2, _2) f(3, _3) f(4, _4)
+#define _ban_fe_6(f, _0, _1, _2, _3, _4, _5)             f(0, _0) f(1, _1) f(2, _2) f(3, _3) f(4, _4) f(5, _5)
+#define _ban_fe_7(f, _0, _1, _2, _3, _4, _5, _6)         f(0, _0) f(1, _1) f(2, _2) f(3, _3) f(4, _4) f(5, _5) f(6, _6)
+#define _ban_fe_8(f, _0, _1, _2, _3, _4, _5, _6, _7)     f(0, _0) f(1, _1) f(2, _2) f(3, _3) f(4, _4) f(5, _5) f(6, _6) f(7, _7)
+#define _ban_fe_9(f, _0, _1, _2, _3, _4, _5, _6, _7, _8) f(0, _0) f(1, _1) f(2, _2) f(3, _3) f(4, _4) f(5, _5) f(6, _6) f(7, _7) f(8, _8)
+#define _ban_for_each(f, ...) _ban_concat(_ban_fe_, _ban_count_args(__VA_ARGS__))(f __VA_OPT__(,) __VA_ARGS__)
+
+#define _ban_fe_comma_0(f)
+#define _ban_fe_comma_1(f, _0)                                 f(0, _0)
+#define _ban_fe_comma_2(f, _0, _1)                             f(0, _0), f(1, _1)
+#define _ban_fe_comma_3(f, _0, _1, _2)                         f(0, _0), f(1, _1), f(2, _2)
+#define _ban_fe_comma_4(f, _0, _1, _2, _3)                     f(0, _0), f(1, _1), f(2, _2), f(3, _3)
+#define _ban_fe_comma_5(f, _0, _1, _2, _3, _4)                 f(0, _0), f(1, _1), f(2, _2), f(3, _3), f(4, _4)
+#define _ban_fe_comma_6(f, _0, _1, _2, _3, _4, _5)             f(0, _0), f(1, _1), f(2, _2), f(3, _3), f(4, _4), f(5, _5)
+#define _ban_fe_comma_7(f, _0, _1, _2, _3, _4, _5, _6)         f(0, _0), f(1, _1), f(2, _2), f(3, _3), f(4, _4), f(5, _5), f(6, _6)
+#define _ban_fe_comma_8(f, _0, _1, _2, _3, _4, _5, _6, _7)     f(0, _0), f(1, _1), f(2, _2), f(3, _3), f(4, _4), f(5, _5), f(6, _6), f(7, _7)
+#define _ban_fe_comma_9(f, _0, _1, _2, _3, _4, _5, _6, _7, _8) f(0, _0), f(1, _1), f(2, _2), f(3, _3), f(4, _4), f(5, _5), f(6, _6), f(7, _7), f(8, _8)
+#define _ban_for_each_comma(f, ...) _ban_concat(_ban_fe_comma_, _ban_count_args(__VA_ARGS__))(f __VA_OPT__(,) __VA_ARGS__)
+
+#define _ban_get_0(a0, ...)                                     a0
+#define _ban_get_1(a0, a1, ...)                                 a1
+#define _ban_get_2(a0, a1, a2, ...)                             a2
+#define _ban_get_3(a0, a1, a2, a3, ...)                         a3
+#define _ban_get_4(a0, a1, a2, a3, a4, ...)                     a4
+#define _ban_get_5(a0, a1, a2, a3, a4, a5, ...)                 a5
+#define _ban_get_6(a0, a1, a2, a3, a4, a5, a6, ...)             a6
+#define _ban_get_7(a0, a1, a2, a3, a4, a5, a6, a7, ...)         a7
+#define _ban_get_8(a0, a1, a2, a3, a4, a5, a6, a7, a8, ...)     a8
+#define _ban_get_9(a0, a1, a2, a3, a4, a5, a6, a7, a8, a9, ...) a9
+#define _ban_get(n, ...) _ban_concat(_ban_get_, n)(__VA_ARGS__)

BAN/include/BAN/Math.h

@@ -36,12 +36,11 @@ namespace BAN::Math
 	template<integral T>
 	inline constexpr T gcd(T a, T b)
 	{
-		T t;
 		while (b)
 		{
-			t = b;
+			T temp = b;
 			b = a % b;
-			a = t;
+			a = temp;
 		}
 		return a;
 	}
@@ -66,25 +65,20 @@ namespace BAN::Math
 		return (x & (x - 1)) == 0;
 	}
 
-	template<BAN::integral T>
+	template<integral T>
-	static constexpr bool will_multiplication_overflow(T a, T b)
+	__attribute__((always_inline))
+	inline constexpr bool will_multiplication_overflow(T a, T b)
 	{
-		if (a == 0 || b == 0)
+		T dummy;
-			return false;
+		return __builtin_mul_overflow(a, b, &dummy);
-		if ((a > 0) == (b > 0))
-			return a > BAN::numeric_limits<T>::max() / b;
-		else
-			return a < BAN::numeric_limits<T>::min() / b;
 	}
 
-	template<BAN::integral T>
+	template<integral T>
-	static constexpr bool will_addition_overflow(T a, T b)
+	__attribute__((always_inline))
+	inline constexpr bool will_addition_overflow(T a, T b)
 	{
-		if (a > 0 && b > 0)
+		T dummy;
-			return a > BAN::numeric_limits<T>::max() - b;
+		return __builtin_add_overflow(a, b, &dummy);
-		if (a < 0 && b < 0)
-			return a < BAN::numeric_limits<T>::min() - b;
-		return false;
 	}
 
 	template<typename T>
@@ -98,6 +92,19 @@ namespace BAN::Math
 		return sizeof(T) * 8 - __builtin_clzll(x) - 1;
 	}
 
+// This is ugly but my clangd does not like including
+// intrinsic headers at all
+#if !defined(__SSE__) || !defined(__SSE2__)
+#pragma GCC push_options
+#ifndef __SSE__
+#pragma GCC target("sse")
+#endif
+#ifndef __SSE2__
+#pragma GCC target("sse2")
+#endif
+#define BAN_MATH_POP_OPTIONS
+#endif
+
 	template<floating_point T>
 	inline constexpr T floor(T x)
 	{
@@ -159,7 +166,23 @@ namespace BAN::Math
 			"jne 1b;"
 			: "+t"(a)
 			: "u"(b)
-			: "ax"
+			: "ax", "cc"
+		);
+		return a;
+	}
+
+	template<floating_point T>
+	inline constexpr T remainder(T a, T b)
+	{
+		asm(
+			"1:"
+			"fprem1;"
+			"fnstsw %%ax;"
+			"testb $4, %%ah;"
+			"jne 1b;"
+			: "+t"(a)
+			: "u"(b)
+			: "ax", "cc"
 		);
 		return a;
 	}
@@ -167,7 +190,7 @@ namespace BAN::Math
 	template<floating_point T>
 	static T modf(T x, T* iptr)
 	{
-		const T frac = BAN::Math::fmod<T>(x, 1);
+		const T frac = BAN::Math::fmod<T>(x, (T)1.0);
 		*iptr = x - frac;
 		return frac;
 	}
@@ -175,15 +198,15 @@ namespace BAN::Math
 	template<floating_point T>
 	inline constexpr T frexp(T num, int* exp)
 	{
-		if (num == 0.0)
+		if (num == (T)0.0)
 		{
 			*exp = 0;
-			return 0.0;
+			return (T)0.0;
 		}
 
-		T _exp;
+		T e;
-		asm("fxtract" : "+t"(num), "=u"(_exp));
+		asm("fxtract" : "+t"(num), "=u"(e));
-		*exp = (int)_exp + 1;
+		*exp = (int)e + 1;
 		return num / (T)2.0;
 	}
 
@@ -251,6 +274,7 @@ namespace BAN::Math
 			"fstp %%st(1);"
 			: "+t"(x)
 		);
+
 		return x;
 	}
 
@@ -263,18 +287,9 @@ namespace BAN::Math
 	template<floating_point T>
 	inline constexpr T pow(T x, T y)
 	{
-		asm(
+		if (x == (T)0.0)
-			"fyl2x;"
+			return (T)0.0;
-			"fld1;"
+		return exp2<T>(y * log2<T>(x));
-			"fld %%st(1);"
-			"fprem;"
-			"f2xm1;"
-			"faddp;"
-			"fscale;"
-			: "+t"(x), "+u"(y)
-		);
-
-		return x;
 	}
 
 	template<floating_point T>
@@ -310,16 +325,27 @@ namespace BAN::Math
 	template<floating_point T>
 	inline constexpr T sqrt(T x)
 	{
-		asm("fsqrt" : "+t"(x));
+		if constexpr(BAN::is_same_v<T, float>)
-		return x;
+		{
+			using v4sf = float __attribute__((vector_size(16)));
+			return __builtin_ia32_sqrtss((v4sf) { x, 0.0f, 0.0f, 0.0f })[0];
+		}
+		else if constexpr(BAN::is_same_v<T, double>)
+		{
+			using v2df = double __attribute__((vector_size(16)));
+			return __builtin_ia32_sqrtsd((v2df) { x, 0.0 })[0];
+		}
+		else if constexpr(BAN::is_same_v<T, long double>)
+		{
+			asm("fsqrt" : "+t"(x));
+			return x;
+		}
 	}
 
 	template<floating_point T>
 	inline constexpr T cbrt(T value)
 	{
-		if (value == 0.0)
+		return pow<T>(value, (T)1.0 / (T)3.0);
-			return 0.0;
-		return pow<T>(value, 1.0 / 3.0);
 	}
 
 	template<floating_point T>
@@ -346,30 +372,21 @@ namespace BAN::Math
 	inline constexpr T tan(T x)
 	{
 		T one, ret;
-		asm(
+		asm("fptan" : "=t"(one), "=u"(ret) : "0"(x));
-			"fptan"
-			: "=t"(one), "=u"(ret)
-			: "0"(x)
-		);
 		return ret;
 	}
 
 	template<floating_point T>
 	inline constexpr T atan2(T y, T x)
 	{
-		asm(
+		asm("fpatan" : "+t"(x) : "u"(y) : "st(1)");
-			"fpatan"
-			: "+t"(x)
-			: "u"(y)
-			: "st(1)"
-		);
 		return x;
 	}
 
 	template<floating_point T>
 	inline constexpr T atan(T x)
 	{
-		return atan2<T>(x, 1.0);
+		return atan2<T>(x, (T)1.0);
 	}
 
 	template<floating_point T>
@@ -378,10 +395,10 @@ namespace BAN::Math
 		if (x == (T)0.0)
 			return (T)0.0;
 		if (x == (T)1.0)
-			return  numbers::pi_v<T> / (T)2.0;
+			return +numbers::pi_v<T> / (T)2.0;
 		if (x == (T)-1.0)
 			return -numbers::pi_v<T> / (T)2.0;
-		return (T)2.0 * atan<T>(x / (T(1.0) + sqrt<T>((T)1.0 - x * x)));
+		return (T)2.0 * atan<T>(x / ((T)1.0 + sqrt<T>((T)1.0 - x * x)));
 	}
 
 	template<floating_point T>
@@ -411,7 +428,7 @@ namespace BAN::Math
 	template<floating_point T>
 	inline constexpr T tanh(T x)
 	{
-		const T exp_px = exp<T>(x);
+		const T exp_px = exp<T>(+x);
 		const T exp_nx = exp<T>(-x);
 		return (exp_px - exp_nx) / (exp_px + exp_nx);
 	}
@@ -440,4 +457,9 @@ namespace BAN::Math
 		return sqrt<T>(x * x + y * y);
 	}
 
+#ifdef BAN_MATH_POP_OPTIONS
+#undef BAN_MATH_POP_OPTIONS
+#pragma GCC pop_options
+#endif
+
 }

BAN/include/BAN/New.h

@@ -9,10 +9,12 @@
 namespace BAN
 {
 	#if defined(__is_kernel)
-		static constexpr void*(&allocator)(size_t) = kmalloc;
+		static constexpr void*(*allocator)(size_t) = kmalloc;
-		static constexpr void(&deallocator)(void*) = kfree;
+		static constexpr void*(*reallocator)(void*, size_t) = nullptr;
+		static constexpr void(*deallocator)(void*) = kfree;
 	#else
-		static constexpr void*(&allocator)(size_t) = malloc;
+		static constexpr void*(*allocator)(size_t) = malloc;
-		static constexpr void(&deallocator)(void*) = free;
+		static constexpr void*(*reallocator)(void*, size_t) = realloc;
+		static constexpr void(*deallocator)(void*) = free;
 	#endif
 }

BAN/include/BAN/PriorityQueue.h

@@ -0,0 +1,64 @@
+#pragma once
+
+#include "BAN/Errors.h"
+#include <BAN/Vector.h>
+#include <BAN/Heap.h>
+
+namespace BAN
+{
+
+	template<typename T, typename Comp = less<T>>
+	class PriorityQueue
+	{
+	public:
+		PriorityQueue() = default;
+		PriorityQueue(Comp comp)
+			: m_comp(comp)
+		{ }
+
+		ErrorOr<void> push(const T& value)
+		{
+			TRY(m_data.push_back(value));
+			push_heap(m_data.begin(), m_data.end());
+			return {};
+		}
+
+		ErrorOr<void> push(T&& value)
+		{
+			TRY(m_data.push_back(move(value)));
+			push_heap(m_data.begin(), m_data.end());
+			return {};
+		}
+
+		template<typename... Args>
+		ErrorOr<void> emplace(Args&&... args) requires is_constructible_v<T, Args...>
+		{
+			TRY(m_data.emplace_back(forward<Args>(args)...));
+			push_heap(m_data.begin(), m_data.end());
+			return {};
+		}
+
+		void pop()
+		{
+			pop_heap(m_data.begin(), m_data.end());
+			m_data.pop_back();
+		}
+
+		BAN::ErrorOr<void> reserve(Vector<T>::size_type size)
+		{
+			return m_data.reserve(size);
+		}
+
+		T& top() { return m_data.front(); }
+		const T& top() const { return m_data.front(); }
+
+		bool empty() const { return m_data.empty(); }
+		Vector<T>::size_type size() const { return m_data.size(); }
+		Vector<T>::size_type capacity() const { return m_data.capacity(); }
+
+	private:
+		Comp m_comp;
+		Vector<T> m_data;
+	};
+
+}

BAN/include/BAN/Sort.h

@@ -1,5 +1,6 @@
 #pragma once
 
+#include <BAN/Heap.h>
 #include <BAN/Math.h>
 #include <BAN/Swap.h>
 #include <BAN/Traits.h>
@@ -8,7 +9,7 @@
 namespace BAN::sort
 {
 
-	template<typename It, typename Comp = less<typename It::value_type>>
+	template<typename It, typename Comp = less<it_value_type_t<It>>>
 	void exchange_sort(It begin, It end, Comp comp = {})
 	{
 		for (It lhs = begin; lhs != end; ++lhs)
@@ -42,7 +43,7 @@ namespace BAN::sort
 
 	}
 
-	template<typename It, typename Comp = less<typename It::value_type>>
+	template<typename It, typename Comp = less<it_value_type_t<It>>>
 	void quick_sort(It begin, It end, Comp comp = {})
 	{
 		if (distance(begin, end) <= 1)
@@ -52,14 +53,14 @@ namespace BAN::sort
 		quick_sort(++mid, end, comp);
 	}
 
-	template<typename It, typename Comp = less<typename It::value_type>>
+	template<typename It, typename Comp = less<it_value_type_t<It>>>
 	void insertion_sort(It begin, It end, Comp comp = {})
 	{
 		if (distance(begin, end) <= 1)
 			return;
 		for (It it1 = next(begin, 1); it1 != end; ++it1)
 		{
-			typename It::value_type x = move(*it1);
+			auto x = move(*it1);
 			It it2 = it1;
 			for (; it2 != begin && comp(x, *prev(it2, 1)); --it2)
 				*it2 = move(*prev(it2, 1));
@@ -67,83 +68,7 @@ namespace BAN::sort
 		}
 	}
 
-	namespace detail
+	template<typename It, typename Comp = less<it_value_type_t<It>>>
-	{
-
-		template<typename It, typename Comp>
-		void push_heap(It begin, size_t hole_index, size_t top_index, typename It::value_type value, Comp comp)
-		{
-			size_t parent = (hole_index - 1) / 2;
-			while (hole_index > top_index && comp(*next(begin, parent), value))
-			{
-				*next(begin, hole_index) = move(*next(begin, parent));
-				hole_index = parent;
-				parent = (hole_index - 1) / 2;
-			}
-			*next(begin, hole_index) = move(value);
-		}
-
-		template<typename It, typename Comp>
-		void adjust_heap(It begin, size_t hole_index, size_t len, typename It::value_type value, Comp comp)
-		{
-			const size_t top_index = hole_index;
-			size_t child = hole_index;
-			while (child < (len - 1) / 2)
-			{
-				child = 2 * (child + 1);
-				if (comp(*next(begin, child), *next(begin, child - 1)))
-					child--;
-				*next(begin, hole_index) = move(*next(begin, child));
-				hole_index = child;
-			}
-			if (len % 2 == 0 && child == (len - 2) / 2)
-			{
-				child = 2 * (child + 1);
-				*next(begin, hole_index) = move(*next(begin, child - 1));
-				hole_index = child - 1;
-			}
-			push_heap(begin, hole_index, top_index, move(value), comp);
-		}
-
-	}
-
-	template<typename It, typename Comp = less<typename It::value_type>>
-	void make_heap(It begin, It end, Comp comp = {})
-	{
-		const size_t len = distance(begin, end);
-		if (len <= 1)
-			return;
-
-		size_t parent = (len - 2) / 2;
-		while (true)
-		{
-			detail::adjust_heap(begin, parent, len, move(*next(begin, parent)), comp);
-
-			if (parent == 0)
-				break;
-
-			parent--;
-		}
-	}
-
-	template<typename It, typename Comp = less<typename It::value_type>>
-	void sort_heap(It begin, It end, Comp comp = {})
-	{
-		const size_t len = distance(begin, end);
-		if (len <= 1)
-			return;
-
-		size_t last = len;
-		while (last > 1)
-		{
-			last--;
-			typename It::value_type x = move(*next(begin, last));
-			*next(begin, last) = move(*begin);
-			detail::adjust_heap(begin, 0, last, move(x), comp);
-		}
-	}
-
-	template<typename It, typename Comp = less<typename It::value_type>>
 	void heap_sort(It begin, It end, Comp comp = {})
 	{
 		make_heap(begin, end, comp);
@@ -167,7 +92,7 @@ namespace BAN::sort
 
 	}
 
-	template<typename It, typename Comp = less<typename It::value_type>>
+	template<typename It, typename Comp = less<it_value_type_t<It>>>
 	void intro_sort(It begin, It end, Comp comp = {})
 	{
 		const size_t len = distance(begin, end);
@@ -190,10 +115,10 @@ namespace BAN::sort
 	}
 
 	template<typename It, size_t radix = 256>
-	requires is_unsigned_v<typename It::value_type> && (radix > 0 && (radix & (radix - 1)) == 0)
+	requires is_unsigned_v<it_value_type_t<It>> && (radix > 0 && (radix & (radix - 1)) == 0)
 	BAN::ErrorOr<void> radix_sort(It begin, It end)
 	{
-		using value_type = typename It::value_type;
+		using value_type = it_value_type_t<It>;
 
 		const size_t len = distance(begin, end);
 		if (len <= 1)
@@ -231,7 +156,7 @@ namespace BAN::sort
 		return {};
 	}
 
-	template<typename It, typename Comp = less<typename It::value_type>>
+	template<typename It, typename Comp = less<it_value_type_t<It>>>
 	void sort(It begin, It end, Comp comp = {})
 	{
 		return intro_sort(begin, end, comp);

BAN/include/BAN/Span.h

@@ -69,7 +69,6 @@ namespace BAN
 
 		value_type* data() const
 		{
-			ASSERT(m_data);
 			return m_data;
 		}
 
@@ -84,7 +83,6 @@ namespace BAN
 
 		Span slice(size_type start, size_type length = ~size_type(0)) const
 		{
-			ASSERT(m_data);
 			ASSERT(start <= m_size);
 			if (length == ~size_type(0))
 				length = m_size - start;

BAN/include/BAN/String.h

@@ -14,6 +14,7 @@ namespace BAN
 	{
 	public:
 		using size_type = size_t;
+		using value_type = char;
 		using iterator = IteratorSimple<char, String>;
 		using const_iterator = ConstIteratorSimple<char, String>;
 		static constexpr size_type sso_capacity = 15;
@@ -352,10 +353,9 @@ namespace BAN::Formatter
 {
 
 	template<typename F>
-	void print_argument(F putc, const String& string, const ValueFormat&)
+	void print_argument(F putc, const String& string, const ValueFormat& format)
 	{
-		for (String::size_type i = 0; i < string.size(); i++)
+		print_argument(putc, string.sv(), format);
-			putc(string[i]);
 	}
 
 }

BAN/include/BAN/StringView.h

@@ -14,6 +14,7 @@ namespace BAN
 	{
 	public:
 		using size_type = size_t;
+		using value_type = char;
 		using const_iterator = ConstIteratorSimple<char, StringView>;
 
 	public:
@@ -246,10 +247,12 @@ namespace BAN::Formatter
 {
 
 	template<typename F>
-	void print_argument(F putc, const StringView& sv, const ValueFormat&)
+	void print_argument(F putc, const StringView& sv, const ValueFormat& format)
 	{
 		for (StringView::size_type i = 0; i < sv.size(); i++)
 			putc(sv[i]);
+		for (int i = sv.size(); i < format.fill; i++)
+			putc(' ');
 	}
 
 }

BAN/include/BAN/Traits.h

@@ -61,6 +61,9 @@ namespace BAN
 	template<typename T> struct is_move_constructible { static constexpr bool value = is_constructible_v<T, T&&>; };
 	template<typename T> inline constexpr bool is_move_constructible_v = is_move_constructible<T>::value;
 
+	template<typename T> struct is_trivially_copyable { static constexpr bool value = __is_trivially_copyable(T); };
+	template<typename T> inline constexpr bool is_trivially_copyable_v = is_trivially_copyable<T>::value;
+
 	template<typename T> struct is_integral { static constexpr bool value = requires (T t, T* p, void (*f)(T)) { reinterpret_cast<T>(t); f(0); p + t; }; };
 	template<typename T> inline constexpr bool is_integral_v = is_integral<T>::value;
 	template<typename T> concept integral = is_integral_v<T>;
@@ -139,6 +142,10 @@ namespace BAN
 	template<typename T> using make_signed_t = typename make_signed<T>::type;
 #undef __BAN_TRAITS_MAKE_SIGNED_CV
 
+	template<typename T> struct it_value_type { using value_type = T::value_type; };
+	template<typename T> struct it_value_type<T*> { using value_type = T; };
+	template<typename T> using it_value_type_t = typename it_value_type<T>::value_type;
+
 	template<typename T> struct less	{ constexpr bool operator()(const T& lhs, const T& rhs) const { return lhs < rhs; } };
 	template<typename T> struct equal	{ constexpr bool operator()(const T& lhs, const T& rhs) const { return lhs == rhs; } };
 	template<typename T> struct greater	{ constexpr bool operator()(const T& lhs, const T& rhs) const { return lhs > rhs; } };

BAN/include/BAN/Variant.h

@@ -126,14 +126,16 @@ namespace BAN
 		Variant(Variant&& other)
 			: m_index(other.m_index)
 		{
-			detail::move_construct<Ts...>(other.m_index, other.m_storage, m_storage);
+			if (other.has_value())
+				detail::move_construct<Ts...>(other.m_index, other.m_storage, m_storage);
 			other.clear();
 		}
 
 		Variant(const Variant& other)
 			: m_index(other.m_index)
 		{
-			detail::copy_construct<Ts...>(other.m_index, other.m_storage, m_storage);
+			if (other.has_value())
+				detail::copy_construct<Ts...>(other.m_index, other.m_storage, m_storage);
 		}
 
 		template<typename T>
@@ -157,12 +159,13 @@ namespace BAN
 
 		Variant& operator=(Variant&& other)
 		{
-			if (m_index == other.m_index)
+			if (m_index == other.m_index && m_index != invalid_index())
 				detail::move_assign<Ts...>(m_index, other.m_storage, m_storage);
 			else
 			{
 				clear();
-				detail::move_construct<Ts...>(other.m_index, other.m_storage, m_storage);
+				if (other.has_value())
+					detail::move_construct<Ts...>(other.m_index, other.m_storage, m_storage);
 				m_index = other.m_index;
 			}
 			other.clear();
@@ -171,12 +174,13 @@ namespace BAN
 
 		Variant& operator=(const Variant& other)
 		{
-			if (m_index == other.m_index)
+			if (m_index == other.m_index && m_index != invalid_index())
 				detail::copy_assign<Ts...>(m_index, other.m_storage, m_storage);
 			else
 			{
 				clear();
-				detail::copy_construct<Ts...>(other.m_index, other.m_storage, m_storage);
+				if (other.has_value())
+					detail::copy_construct<Ts...>(other.m_index, other.m_storage, m_storage);
 				m_index = other.m_index;
 			}
 			return *this;

BAN/include/BAN/Vector.h

@@ -56,7 +56,7 @@ namespace BAN
 		bool contains(const T&) const;
 
 		Span<T> span() { return Span(m_data, m_size); }
-		const Span<T> span() const { return Span(m_data, m_size); }
+		Span<const T> span() const { return Span(m_data, m_size); }
 
 		const T& operator[](size_type) const;
 		T& operator[](size_type);
@@ -381,19 +381,46 @@ namespace BAN
 	template<typename T>
 	ErrorOr<void> Vector<T>::ensure_capacity(size_type size)
 	{
+		static_assert(alignof(T) <= alignof(max_align_t), "over aligned types not supported");
+
 		if (m_capacity >= size)
 			return {};
-		size_type new_cap = BAN::Math::max<size_type>(size, m_capacity * 2);
+
-		T* new_data = (T*)BAN::allocator(new_cap * sizeof(T));
+		const size_type new_cap = BAN::Math::max<size_type>(size, m_capacity * 2);
-		if (new_data == nullptr)
+
-			return Error::from_errno(ENOMEM);
+		if constexpr (BAN::is_trivially_copyable_v<T>)
-		for (size_type i = 0; i < m_size; i++)
 		{
-			new (new_data + i) T(move(m_data[i]));
+			if constexpr (BAN::reallocator)
-			m_data[i].~T();
+			{
+				auto* new_data = static_cast<T*>(BAN::reallocator(m_data, new_cap * sizeof(T)));
+				if (new_data == nullptr)
+					return Error::from_errno(ENOMEM);
+				m_data = new_data;
+			}
+			else
+			{
+				auto* new_data = static_cast<T*>(BAN::allocator(new_cap * sizeof(T)));
+				if (new_data == nullptr)
+					return Error::from_errno(ENOMEM);
+				memcpy(new_data, m_data, m_size * sizeof(T));
+				BAN::deallocator(m_data);
+				m_data = new_data;
+			}
 		}
-		BAN::deallocator(m_data);
+		else
-		m_data = new_data;
+		{
+			auto* new_data = static_cast<T*>(BAN::allocator(new_cap * sizeof(T)));
+			if (new_data == nullptr)
+				return Error::from_errno(ENOMEM);
+			for (size_type i = 0; i < m_size; i++)
+			{
+				new (new_data + i) T(move(m_data[i]));
+				m_data[i].~T();
+			}
+			BAN::deallocator(m_data);
+			m_data = new_data;
+		}
+
 		m_capacity = new_cap;
 		return {};
 	}

kernel/CMakeLists.txt

@@ -5,9 +5,12 @@ set(KERNEL_SOURCES
 	kernel/ACPI/AML/Node.cpp
 	kernel/ACPI/AML/OpRegion.cpp
 	kernel/ACPI/BatterySystem.cpp
+	kernel/ACPI/EmbeddedController.cpp
 	kernel/APIC.cpp
 	kernel/Audio/AC97/Controller.cpp
 	kernel/Audio/Controller.cpp
+	kernel/Audio/HDAudio/AudioFunctionGroup.cpp
+	kernel/Audio/HDAudio/Controller.cpp
 	kernel/BootInfo.cpp
 	kernel/CPUID.cpp
 	kernel/Credentials.cpp
@@ -22,6 +25,7 @@ set(KERNEL_SOURCES
 	kernel/Epoll.cpp
 	kernel/Errors.cpp
 	kernel/FS/DevFS/FileSystem.cpp
+	kernel/FS/EventFD.cpp
 	kernel/FS/Ext2/FileSystem.cpp
 	kernel/FS/Ext2/Inode.cpp
 	kernel/FS/FAT/FileSystem.cpp
@@ -47,6 +51,7 @@ set(KERNEL_SOURCES
 	kernel/InterruptController.cpp
 	kernel/kernel.cpp
 	kernel/Lock/SpinLock.cpp
+	kernel/Memory/ByteRingBuffer.cpp
 	kernel/Memory/DMARegion.cpp
 	kernel/Memory/FileBackedRegion.cpp
 	kernel/Memory/Heap.cpp
@@ -107,6 +112,7 @@ set(KERNEL_SOURCES
 	kernel/USB/Controller.cpp
 	kernel/USB/Device.cpp
 	kernel/USB/HID/HIDDriver.cpp
+	kernel/USB/HID/Joystick.cpp
 	kernel/USB/HID/Keyboard.cpp
 	kernel/USB/HID/Mouse.cpp
 	kernel/USB/Hub/HubDriver.cpp
@@ -133,6 +139,8 @@ if("${BANAN_ARCH}" STREQUAL "x86_64")
 		arch/x86_64/Signal.S
 		arch/x86_64/Syscall.S
 		arch/x86_64/Thread.S
+		arch/x86_64/User.S
+		arch/x86_64/Yield.S
 	)
 elseif("${BANAN_ARCH}" STREQUAL "i686")
 	set(KERNEL_SOURCES
@@ -143,6 +151,8 @@ elseif("${BANAN_ARCH}" STREQUAL "i686")
 		arch/i686/Signal.S
 		arch/i686/Syscall.S
 		arch/i686/Thread.S
+		arch/i686/User.S
+		arch/i686/Yield.S
 	)
 else()
 	message(FATAL_ERROR "unsupported architecure ${BANAN_ARCH}")
@@ -158,10 +168,13 @@ set(BAN_SOURCES
 set(KLIBC_SOURCES
 	klibc/ctype.cpp
 	klibc/string.cpp
+	klibc/arch/${BANAN_ARCH}/string.S
+)
 
-	# Ehhh don't do this but for now libc uses the same stuff kernel can use
+set(LIBDEFLATE_SOURCE
-	# This won't work after libc starts using sse implemetations tho
+	../userspace/libraries/LibDEFLATE/Compressor.cpp
-	../userspace/libraries/LibC/arch/${BANAN_ARCH}/string.S
+	../userspace/libraries/LibDEFLATE/Decompressor.cpp
+	../userspace/libraries/LibDEFLATE/HuffmanTree.cpp
 )
 
 set(LIBFONT_SOURCES
@@ -174,18 +187,25 @@ set(LIBINPUT_SOURCE
 	../userspace/libraries/LibInput/KeyEvent.cpp
 )
 
+set(LIBQR_SOURCE
+	../userspace/libraries/LibQR/QRCode.cpp
+)
+
 set(KERNEL_SOURCES
 	${KERNEL_SOURCES}
 	${BAN_SOURCES}
 	${KLIBC_SOURCES}
+	${LIBDEFLATE_SOURCE}
 	${LIBFONT_SOURCES}
 	${LIBINPUT_SOURCE}
+	${LIBQR_SOURCE}
 )
 
 add_executable(kernel ${KERNEL_SOURCES})
 
 target_compile_definitions(kernel PRIVATE __is_kernel)
 target_compile_definitions(kernel PRIVATE __arch=${BANAN_ARCH})
+target_compile_definitions(kernel PRIVATE LIBDEFLATE_AVOID_STACK=1)
 
 target_compile_options(kernel PRIVATE
 	-O2 -g
@@ -239,9 +259,11 @@ add_custom_command(
 
 banan_include_headers(kernel ban)
 banan_include_headers(kernel libc)
-banan_include_headers(kernel libfont)
+banan_include_headers(kernel libdeflate)
 banan_include_headers(kernel libelf)
+banan_include_headers(kernel libfont)
 banan_include_headers(kernel libinput)
+banan_include_headers(kernel libqr)
 
 banan_install_headers(kernel)
 set_target_properties(kernel PROPERTIES OUTPUT_NAME banan-os.kernel)

kernel/arch/i686/PageTable.cpp

@@ -21,6 +21,11 @@ namespace Kernel
 
 	SpinLock PageTable::s_fast_page_lock;
 
+	constexpr uint64_t s_page_flag_mask = 0x8000000000000FFF;
+	constexpr uint64_t s_page_addr_mask = ~s_page_flag_mask;
+
+	static bool s_is_post_heap_done = false;
+
 	static PageTable* s_kernel = nullptr;
 	static bool s_has_nxe = false;
 	static bool s_has_pge = false;
@@ -67,7 +72,7 @@ namespace Kernel
 
 	void PageTable::initialize_post_heap()
 	{
-		// NOTE: this is no-op as our 32 bit target does not use hhdm
+		s_is_post_heap_done = true;
 	}
 
 	void PageTable::initial_load()
@@ -141,9 +146,9 @@ namespace Kernel
 	}
 
 	template<typename T>
-	static vaddr_t P2V(const T paddr)
+	static uint64_t* P2V(const T paddr)
 	{
-		return (paddr_t)paddr - g_boot_info.kernel_paddr + KERNEL_OFFSET;
+		return reinterpret_cast<uint64_t*>(reinterpret_cast<paddr_t>(paddr) - g_boot_info.kernel_paddr + KERNEL_OFFSET);
 	}
 
 	void PageTable::initialize_kernel()
@@ -193,13 +198,18 @@ namespace Kernel
 	{
 		constexpr uint64_t pdpte = (fast_page() >> 30) & 0x1FF;
 		constexpr uint64_t pde   = (fast_page() >> 21) & 0x1FF;
+		constexpr uint64_t pte   = (fast_page() >> 12) & 0x1FF;
 
-		uint64_t* pdpt = reinterpret_cast<uint64_t*>(P2V(m_highest_paging_struct));
+		const uint64_t* pdpt = P2V(m_highest_paging_struct);
 		ASSERT(pdpt[pdpte] & Flags::Present);
 
-		uint64_t* pd = reinterpret_cast<uint64_t*>(P2V(pdpt[pdpte]) & PAGE_ADDR_MASK);
+		uint64_t* pd = P2V(pdpt[pdpte] & s_page_addr_mask);
 		ASSERT(!(pd[pde] & Flags::Present));
 		pd[pde] = V2P(allocate_zeroed_page_aligned_page()) | Flags::ReadWrite | Flags::Present;
+
+		uint64_t* pt = P2V(pd[pde] & s_page_addr_mask);
+		ASSERT(pt[pte] == 0);
+		pt[pte] = Flags::Reserved;
 	}
 
 	void PageTable::map_fast_page(paddr_t paddr)
@@ -214,14 +224,14 @@ namespace Kernel
 		constexpr uint64_t pde   = (fast_page() >> 21) & 0x1FF;
 		constexpr uint64_t pte   = (fast_page() >> 12) & 0x1FF;
 
-		uint64_t* pdpt = reinterpret_cast<uint64_t*>(P2V(s_kernel->m_highest_paging_struct));
+		uint64_t* pdpt = P2V(s_kernel->m_highest_paging_struct);
-		uint64_t* pd   = reinterpret_cast<uint64_t*>(P2V(pdpt[pdpte] & PAGE_ADDR_MASK));
+		uint64_t* pd   = P2V(pdpt[pdpte] & s_page_addr_mask);
-		uint64_t* pt   = reinterpret_cast<uint64_t*>(P2V(pd[pde] & PAGE_ADDR_MASK));
+		uint64_t* pt   = P2V(pd[pde] & s_page_addr_mask);
 
 		ASSERT(!(pt[pte] & Flags::Present));
 		pt[pte] = paddr | Flags::ReadWrite | Flags::Present;
 
-		invalidate(fast_page(), false);
+		asm volatile("invlpg (%0)" :: "r"(fast_page()) : "memory");
 	}
 
 	void PageTable::unmap_fast_page()
@@ -234,14 +244,14 @@ namespace Kernel
 		constexpr uint64_t pde   = (fast_page() >> 21) & 0x1FF;
 		constexpr uint64_t pte   = (fast_page() >> 12) & 0x1FF;
 
-		uint64_t* pdpt = reinterpret_cast<uint64_t*>(P2V(s_kernel->m_highest_paging_struct));
+		uint64_t* pdpt = P2V(s_kernel->m_highest_paging_struct);
-		uint64_t* pd   = reinterpret_cast<uint64_t*>(P2V(pdpt[pdpte] & PAGE_ADDR_MASK));
+		uint64_t* pd   = P2V(pdpt[pdpte] & s_page_addr_mask);
-		uint64_t* pt   = reinterpret_cast<uint64_t*>(P2V(pd[pde] & PAGE_ADDR_MASK));
+		uint64_t* pt   = P2V(pd[pde] & s_page_addr_mask);
 
 		ASSERT(pt[pte] & Flags::Present);
-		pt[pte] = 0;
+		pt[pte] = Flags::Reserved;
 
-		invalidate(fast_page(), false);
+		asm volatile("invlpg (%0)" :: "r"(fast_page()) : "memory");
 	}
 
 	BAN::ErrorOr<PageTable*> PageTable::create_userspace()
@@ -263,7 +273,7 @@ namespace Kernel
 		m_highest_paging_struct = V2P(kmalloc(32, 32, true));
 		ASSERT(m_highest_paging_struct);
 
-		uint64_t* pdpt = reinterpret_cast<uint64_t*>(P2V(m_highest_paging_struct));
+		uint64_t* pdpt = P2V(m_highest_paging_struct);
 		pdpt[0] = 0;
 		pdpt[1] = 0;
 		pdpt[2] = 0;
@@ -276,18 +286,17 @@ namespace Kernel
 		if (m_highest_paging_struct == 0)
 			return;
 
-		uint64_t* pdpt = reinterpret_cast<uint64_t*>(P2V(m_highest_paging_struct));
+		uint64_t* pdpt = P2V(m_highest_paging_struct);
-
 		for (uint32_t pdpte = 0; pdpte < 3; pdpte++)
 		{
 			if (!(pdpt[pdpte] & Flags::Present))
 				continue;
-			uint64_t* pd = reinterpret_cast<uint64_t*>(P2V(pdpt[pdpte] & PAGE_ADDR_MASK));
+			uint64_t* pd = P2V(pdpt[pdpte] & s_page_addr_mask);
 			for (uint32_t pde = 0; pde < 512; pde++)
 			{
 				if (!(pd[pde] & Flags::Present))
 					continue;
-				kfree(reinterpret_cast<uint64_t*>(P2V(pd[pde] & PAGE_ADDR_MASK)));
+				kfree(P2V(pd[pde] & s_page_addr_mask));
 			}
 			kfree(pd);
 		}
@@ -298,15 +307,43 @@ namespace Kernel
 	{
 		SpinLockGuard _(m_lock);
 		ASSERT(m_highest_paging_struct < 0x100000000);
-		const uint32_t pdpt_lo = m_highest_paging_struct;
+		asm volatile("movl %0, %%cr3" :: "r"(static_cast<uint32_t>(m_highest_paging_struct)));
-		asm volatile("movl %0, %%cr3" :: "r"(pdpt_lo));
 		Processor::set_current_page_table(this);
 	}
 
-	void PageTable::invalidate(vaddr_t vaddr, bool send_smp_message)
+	void PageTable::invalidate_range(vaddr_t vaddr, size_t pages, bool send_smp_message)
 	{
 		ASSERT(vaddr % PAGE_SIZE == 0);
-		asm volatile("invlpg (%0)" :: "r"(vaddr) : "memory");
+
+		const bool is_userspace = (vaddr < KERNEL_OFFSET);
+		if (is_userspace && this != &PageTable::current())
+			;
+		else if (pages <= 32 || !s_is_post_heap_done)
+		{
+			for (size_t i = 0; i < pages; i++, vaddr += PAGE_SIZE)
+				asm volatile("invlpg (%0)" :: "r"(vaddr));
+		}
+		else if (is_userspace || !s_has_pge)
+		{
+			asm volatile("movl %0, %%cr3" :: "r"(static_cast<uint32_t>(m_highest_paging_struct)));
+		}
+		else
+		{
+			asm volatile(
+				"movl %%cr4, %%eax;"
+
+				"andl $~0x80, %%eax;"
+				"movl %%eax, %%cr4;"
+
+				"movl %0, %%cr3;"
+
+				"orl $0x80, %%eax;"
+				"movl %%eax, %%cr4;"
+				:
+				: "r"(static_cast<uint32_t>(m_highest_paging_struct))
+				: "eax"
+			);
+		}
 
 		if (send_smp_message)
 		{
@@ -314,13 +351,14 @@ namespace Kernel
 				.type = Processor::SMPMessage::Type::FlushTLB,
 				.flush_tlb = {
 					.vaddr      = vaddr,
-					.page_count = 1
+					.page_count = pages,
+					.page_table = vaddr < KERNEL_OFFSET ? this : nullptr,
 				}
 			});
 		}
 	}
 
-	void PageTable::unmap_page(vaddr_t vaddr, bool send_smp_message)
+	void PageTable::unmap_page(vaddr_t vaddr, bool invalidate)
 	{
 		ASSERT(vaddr);
 		ASSERT(vaddr % PAGE_SIZE == 0);
@@ -339,12 +377,16 @@ namespace Kernel
 		if (is_page_free(vaddr))
 			Kernel::panic("trying to unmap unmapped page 0x{H}", vaddr);
 
-		uint64_t* pdpt = reinterpret_cast<uint64_t*>(P2V(m_highest_paging_struct));
+		uint64_t* pdpt = P2V(m_highest_paging_struct);
-		uint64_t* pd   = reinterpret_cast<uint64_t*>(P2V(pdpt[pdpte] & PAGE_ADDR_MASK));
+		uint64_t* pd   = P2V(pdpt[pdpte] & s_page_addr_mask);
-		uint64_t* pt   = reinterpret_cast<uint64_t*>(P2V(pd[pde] & PAGE_ADDR_MASK));
+		uint64_t* pt   = P2V(pd[pde] & s_page_addr_mask);
+
+		const paddr_t old_paddr = pt[pte] & s_page_addr_mask;
 
 		pt[pte] = 0;
-		invalidate(vaddr, send_smp_message);
+
+		if (invalidate && old_paddr != 0)
+			invalidate_page(vaddr, true);
 	}
 
 	void PageTable::unmap_range(vaddr_t vaddr, size_t size)
@@ -356,17 +398,10 @@ namespace Kernel
 		SpinLockGuard _(m_lock);
 		for (vaddr_t page = 0; page < page_count; page++)
 			unmap_page(vaddr + page * PAGE_SIZE, false);
-
+		invalidate_range(vaddr, page_count, true);
-		Processor::broadcast_smp_message({
-			.type = Processor::SMPMessage::Type::FlushTLB,
-			.flush_tlb = {
-				.vaddr      = vaddr,
-				.page_count = page_count
-			}
-		});
 	}
 
-	void PageTable::map_page_at(paddr_t paddr, vaddr_t vaddr, flags_t flags, MemoryType memory_type, bool send_smp_message)
+	void PageTable::map_page_at(paddr_t paddr, vaddr_t vaddr, flags_t flags, MemoryType memory_type, bool invalidate)
 	{
 		ASSERT(vaddr);
 		ASSERT(vaddr != fast_page());
@@ -401,11 +436,11 @@ namespace Kernel
 
 		SpinLockGuard _(m_lock);
 
-		uint64_t* pdpt = reinterpret_cast<uint64_t*>(P2V(m_highest_paging_struct));
+		uint64_t* pdpt = P2V(m_highest_paging_struct);
 		if (!(pdpt[pdpte] & Flags::Present))
 			pdpt[pdpte] = V2P(allocate_zeroed_page_aligned_page()) | Flags::Present;
 
-		uint64_t* pd = reinterpret_cast<uint64_t*>(P2V(pdpt[pdpte] & PAGE_ADDR_MASK));
+		uint64_t* pd = P2V(pdpt[pdpte] & s_page_addr_mask);
 		if ((pd[pde] & uwr_flags) != uwr_flags)
 		{
 			if (!(pd[pde] & Flags::Present))
@@ -416,10 +451,14 @@ namespace Kernel
 		if (!(flags & Flags::Present))
 			uwr_flags &= ~Flags::Present;
 
-		uint64_t* pt = reinterpret_cast<uint64_t*>(P2V(pd[pde] & PAGE_ADDR_MASK));
+		uint64_t* pt = P2V(pd[pde] & s_page_addr_mask);
+
+		const paddr_t old_paddr = pt[pte] & s_page_addr_mask;
+
 		pt[pte] = paddr | uwr_flags | extra_flags;
 
-		invalidate(vaddr, send_smp_message);
+		if (invalidate && old_paddr != 0)
+			invalidate_page(vaddr, true);
 	}
 
 	void PageTable::map_range_at(paddr_t paddr, vaddr_t vaddr, size_t size, flags_t flags, MemoryType memory_type)
@@ -433,14 +472,49 @@ namespace Kernel
 		SpinLockGuard _(m_lock);
 		for (size_t page = 0; page < page_count; page++)
 			map_page_at(paddr + page * PAGE_SIZE, vaddr + page * PAGE_SIZE, flags, memory_type, false);
+		invalidate_range(vaddr, page_count, true);
+	}
 
-		Processor::broadcast_smp_message({
+	void PageTable::remove_writable_from_range(vaddr_t vaddr, size_t size)
-			.type = Processor::SMPMessage::Type::FlushTLB,
+	{
-			.flush_tlb = {
+		ASSERT(vaddr);
-				.vaddr      = vaddr,
+		ASSERT(vaddr % PAGE_SIZE == 0);
-				.page_count = page_count
+
+		uint32_t pdpte = (vaddr >> 30) & 0x1FF;
+		uint32_t pde   = (vaddr >> 21) & 0x1FF;
+		uint32_t pte   = (vaddr >> 12) & 0x1FF;
+
+		const uint32_t e_pdpte = ((vaddr + size - 1) >> 30) & 0x1FF;
+		const uint32_t e_pde   = ((vaddr + size - 1) >> 21) & 0x1FF;
+		const uint32_t e_pte   = ((vaddr + size - 1) >> 12) & 0x1FF;
+
+		SpinLockGuard _(m_lock);
+
+		const uint64_t* pdpt = P2V(m_highest_paging_struct);
+		for (; pdpte <= e_pdpte; pdpte++)
+		{
+			if (!(pdpt[pdpte] & Flags::Present))
+				continue;
+			const uint64_t* pd = P2V(pdpt[pdpte] & s_page_addr_mask);
+			for (; pde < 512; pde++)
+			{
+				if (pdpte == e_pdpte && pde > e_pde)
+					break;
+				if (!(pd[pde] & Flags::ReadWrite))
+					continue;
+				uint64_t* pt = P2V(pd[pde] & s_page_addr_mask);
+				for (; pte < 512; pte++)
+				{
+					if (pdpte == e_pdpte && pde == e_pde && pte > e_pte)
+						break;
+					pt[pte] &= ~static_cast<uint64_t>(Flags::ReadWrite);
+				}
+				pte = 0;
 			}
-		});
+			pde = 0;
+		}
+
+		invalidate_range(vaddr, size / PAGE_SIZE, true);
 	}
 
 	uint64_t PageTable::get_page_data(vaddr_t vaddr) const
@@ -453,15 +527,15 @@ namespace Kernel
 
 		SpinLockGuard _(m_lock);
 
-		uint64_t* pdpt = (uint64_t*)P2V(m_highest_paging_struct);
+		const uint64_t* pdpt = P2V(m_highest_paging_struct);
 		if (!(pdpt[pdpte] & Flags::Present))
 			return 0;
 
-		uint64_t* pd = (uint64_t*)P2V(pdpt[pdpte] & PAGE_ADDR_MASK);
+		const uint64_t* pd = P2V(pdpt[pdpte] & s_page_addr_mask);
 		if (!(pd[pde] & Flags::Present))
 			return 0;
 
-		uint64_t* pt = (uint64_t*)P2V(pd[pde] & PAGE_ADDR_MASK);
+		const uint64_t* pt = P2V(pd[pde] & s_page_addr_mask);
 		if (!(pt[pte] & Flags::Used))
 			return 0;
 
@@ -475,8 +549,7 @@ namespace Kernel
 
 	paddr_t PageTable::physical_address_of(vaddr_t vaddr) const
 	{
-		uint64_t page_data = get_page_data(vaddr);
+		return get_page_data(vaddr) & s_page_addr_mask;
-		return (page_data & PAGE_ADDR_MASK) & ~(1ull << 63);
 	}
 
 	bool PageTable::is_page_free(vaddr_t vaddr) const
@@ -497,13 +570,13 @@ namespace Kernel
 		return true;
 	}
 
-	bool PageTable::reserve_page(vaddr_t vaddr, bool only_free)
+	bool PageTable::reserve_page(vaddr_t vaddr, bool only_free, bool send_smp_message)
 	{
 		SpinLockGuard _(m_lock);
 		ASSERT(vaddr % PAGE_SIZE == 0);
 		if (only_free && !is_page_free(vaddr))
 			return false;
-		map_page_at(0, vaddr, Flags::Reserved);
+		map_page_at(0, vaddr, Flags::Reserved, MemoryType::Normal, send_smp_message);
 		return true;
 	}
 
@@ -517,7 +590,9 @@ namespace Kernel
 		if (only_free && !is_range_free(vaddr, bytes))
 			return false;
 		for (size_t offset = 0; offset < bytes; offset += PAGE_SIZE)
-			reserve_page(vaddr + offset);
+			reserve_page(vaddr + offset, true, false);
+		invalidate_range(vaddr, bytes / PAGE_SIZE, true);
+
 		return true;
 	}
 
@@ -530,48 +605,47 @@ namespace Kernel
 		if (size_t rem = last_address % PAGE_SIZE)
 			last_address -= rem;
 
-		const uint32_t s_pdpte = (first_address >> 30) & 0x1FF;
+		uint32_t pdpte = (first_address >> 30) & 0x1FF;
-		const uint32_t s_pde   = (first_address >> 21) & 0x1FF;
+		uint32_t pde   = (first_address >> 21) & 0x1FF;
-		const uint32_t s_pte   = (first_address >> 12) & 0x1FF;
+		uint32_t pte   = (first_address >> 12) & 0x1FF;
 
-		const uint32_t e_pdpte = (last_address >> 30) & 0x1FF;
+		const uint32_t e_pdpte = ((last_address - 1) >> 30) & 0x1FF;
-		const uint32_t e_pde   = (last_address >> 21) & 0x1FF;
+		const uint32_t e_pde   = ((last_address - 1) >> 21) & 0x1FF;
-		const uint32_t e_pte   = (last_address >> 12) & 0x1FF;
+		const uint32_t e_pte   = ((last_address - 1) >> 12) & 0x1FF;
 
 		SpinLockGuard _(m_lock);
 
 		// Try to find free page that can be mapped without
 		// allocations (page table with unused entries)
-		uint64_t* pdpt = reinterpret_cast<uint64_t*>(P2V(m_highest_paging_struct));
+		const uint64_t* pdpt = P2V(m_highest_paging_struct);
-		for (uint32_t pdpte = s_pdpte; pdpte < 4; pdpte++)
+		for (; pdpte <= e_pdpte; pdpte++)
 		{
-			if (pdpte > e_pdpte)
-				break;
 			if (!(pdpt[pdpte] & Flags::Present))
 				continue;
-			uint64_t* pd = reinterpret_cast<uint64_t*>(P2V(pdpt[pdpte] & PAGE_ADDR_MASK));
+			const uint64_t* pd = P2V(pdpt[pdpte] & s_page_addr_mask);
-			for (uint32_t pde = s_pde; pde < 512; pde++)
+			for (; pde < 512; pde++)
 			{
 				if (pdpte == e_pdpte && pde > e_pde)
 					break;
 				if (!(pd[pde] & Flags::Present))
 					continue;
-				uint64_t* pt = (uint64_t*)P2V(pd[pde] & PAGE_ADDR_MASK);
+				const uint64_t* pt = P2V(pd[pde] & s_page_addr_mask);
-				for (uint32_t pte = s_pte; pte < 512; pte++)
+				for (; pte < 512; pte++)
 				{
-					if (pdpte == e_pdpte && pde == e_pde && pte >= e_pte)
+					if (pdpte == e_pdpte && pde == e_pde && pte > e_pte)
 						break;
-					if (!(pt[pte] & Flags::Used))
+					if (pt[pte] & Flags::Used)
-					{
+						continue;
-						vaddr_t vaddr = 0;
+					vaddr_t vaddr = 0;
-						vaddr |= (vaddr_t)pdpte << 30;
+					vaddr |= (vaddr_t)pdpte << 30;
-						vaddr |= (vaddr_t)pde   << 21;
+					vaddr |= (vaddr_t)pde   << 21;
-						vaddr |= (vaddr_t)pte   << 12;
+					vaddr |= (vaddr_t)pte   << 12;
-						ASSERT(reserve_page(vaddr));
+					ASSERT(reserve_page(vaddr));
-						return vaddr;
+					return vaddr;
-					}
 				}
+				pte = 0;
 			}
+			pde = 0;
 		}
 
 		// Find any free page
@@ -584,7 +658,7 @@ namespace Kernel
 			}
 		}
 
-		ASSERT_NOT_REACHED();
+		return 0;
 	}
 
 	vaddr_t PageTable::reserve_free_contiguous_pages(size_t page_count, vaddr_t first_address, vaddr_t last_address)
@@ -617,7 +691,7 @@ namespace Kernel
 			}
 		}
 
-		ASSERT_NOT_REACHED();
+		return 0;
 	}
 
 	static void dump_range(vaddr_t start, vaddr_t end, PageTable::flags_t flags)
@@ -640,7 +714,7 @@ namespace Kernel
 		flags_t flags = 0;
 		vaddr_t start = 0;
 
-		uint64_t* pdpt = reinterpret_cast<uint64_t*>(P2V(m_highest_paging_struct));
+		const uint64_t* pdpt = P2V(m_highest_paging_struct);
 		for (uint32_t pdpte = 0; pdpte < 4; pdpte++)
 		{
 			if (!(pdpt[pdpte] & Flags::Present))
@@ -649,7 +723,7 @@ namespace Kernel
 				start = 0;
 				continue;
 			}
-			uint64_t* pd = (uint64_t*)P2V(pdpt[pdpte] & PAGE_ADDR_MASK);
+			const uint64_t* pd = P2V(pdpt[pdpte] & s_page_addr_mask);
 			for (uint64_t pde = 0; pde < 512; pde++)
 			{
 				if (!(pd[pde] & Flags::Present))
@@ -658,7 +732,7 @@ namespace Kernel
 					start = 0;
 					continue;
 				}
-				uint64_t* pt = (uint64_t*)P2V(pd[pde] & PAGE_ADDR_MASK);
+				const uint64_t* pt = P2V(pd[pde] & s_page_addr_mask);
 				for (uint64_t pte = 0; pte < 512; pte++)
 				{
 					if (parse_flags(pt[pte]) != flags)

kernel/arch/i686/Signal.S

@@ -1,16 +1,41 @@
 .section .userspace, "ax"
 
 // stack contains
-//    return address
+//    (4 bytes)  return address (on return stack)
-//    signal number
+//    (4 bytes)  return stack
-//    signal handler
+//    (4 bytes)  return rflags
+//    (8 bytes)  restore sigmask
+//    (36 bytes) siginfo_t
+//    (4 bytes)  signal number
+//    (4 bytes)  signal handler
 
 .global signal_trampoline
 signal_trampoline:
-	pusha
+	pushl %esi // gregs
+	pushl %edi
+	pushl %edx
+	pushl %ecx
+	pushl %ebx
+	pushl %eax
+	pushl %ebp
 
-	movl 40(%esp), %edi
+	movl 80(%esp), %eax
-	movl 36(%esp), %eax
+	pushl %eax; addl $4, (%esp)
+	pushl (%eax)
+
+	// FIXME: populate these
+	xorl %eax, %eax
+	pushl %eax // stack
+	pushl %eax
+	pushl %eax
+	pushl %eax // sigset
+	pushl %eax
+	pushl %eax // link
+
+	movl    %esp,  %edx // ucontext
+	leal 68(%esp), %esi // siginfo
+	movl 64(%esp), %edi // signal number
+	movl 60(%esp), %eax // handlers
 
 	// align stack to 16 bytes
 	movl %esp, %ebp
@@ -19,7 +44,9 @@ signal_trampoline:
 	subl $512, %esp
 	fxsave (%esp)
 
-	subl $12, %esp
+	subl $4, %esp
+	pushl %edx
+	pushl %esi
 	pushl %edi
 	call *%eax
 	addl $16, %esp
@@ -29,9 +56,31 @@ signal_trampoline:
 
 	// restore stack
 	movl %ebp, %esp
-	popa
+	addl $24, %esp
 
+	// restore sigmask
+	movl $83, %eax      // SYS_SIGPROCMASK
+	movl $3,  %ebx      // SIG_SETMASK
+	leal 72(%esp), %ecx // set
+	xorl %edx, %edx     // oset
+	int $0xF0
+
+	// restore registers
 	addl $8, %esp
+	popl %ebp
+	popl %eax
+	popl %ebx
+	popl %ecx
+	popl %edx
+	popl %edi
+	popl %esi
+
+	// skip handler, number, siginfo_t, sigmask
+	addl $52, %esp
+
+	// restore flags
 	popf
 
+	movl (%esp), %esp
+
 	ret

kernel/arch/i686/Syscall.S

@@ -1,12 +1,6 @@
 // arguments in EAX, EBX, ECX, EDX, ESI, EDI
 .global asm_syscall_handler
 asm_syscall_handler:
-	# save segment registers
-	pushw %ds
-	pushw %es
-	pushw %fs
-	pushw %gs
-
 	# save general purpose registers
 	pushl %ebx
 	pushl %ecx
@@ -18,13 +12,10 @@ asm_syscall_handler:
 
 	# align stack
 	movl %esp, %ebp
-	subl $15, %esp
+	andl $-16, %esp
-	andl $0xFFFFFFF0, %esp
 
 	# push arguments
-	subl $4, %esp
+	subl $8, %esp
-	pushl %ebp
-	addl $32, (%esp)
 	pushl %edi
 	pushl %esi
 	pushl %edx
@@ -44,6 +35,15 @@ asm_syscall_handler:
 
 	movl %ebp, %esp
 
+	# restore userspace segments
+	movw $(0x20 | 3), %bx
+	movw %bx, %ds
+	movw %bx, %es
+	movw $(0x30 | 3), %bx
+	movw %bx, %fs
+	movw $(0x38 | 3), %bx
+	movw %bx, %gs
+
 	# restore general purpose registers
 	popl %ebp
 	popl %esi
@@ -52,12 +52,6 @@ asm_syscall_handler:
 	popl %ecx
 	popl %ebx
 
-	# restore segment registers
-	popw %gs
-	popw %fs
-	popw %es
-	popw %ds
-
 	iret
 
 .global sys_fork_trampoline
@@ -69,7 +63,7 @@ sys_fork_trampoline:
 
 	call read_ip
 	testl %eax, %eax
-	jz .reload_stack
+	jz .done
 
 	movl %esp, %ebx
 
@@ -85,9 +79,3 @@ sys_fork_trampoline:
 	popl %ebx
 	popl %ebp
 	ret
-
- .reload_stack:
-	call get_thread_start_sp
-	movl %eax, %esp
-	xorl %eax, %eax
-	jmp .done

kernel/arch/i686/Thread.S

@@ -7,9 +7,6 @@ read_ip:
 # void start_kernel_thread()
 .global start_kernel_thread
 start_kernel_thread:
-	call get_thread_start_sp
-	movl %eax, %esp
-
 	# STACK LAYOUT
 	#   on_exit arg
 	#   on_exit func
@@ -31,25 +28,15 @@ start_kernel_thread:
 	subl $12, %esp
 	pushl %edi
 	call *%esi
-	addl $16, %esp
-
 
 .global start_userspace_thread
 start_userspace_thread:
-	call load_thread_sse
-
-	call get_thread_start_sp
-	movl %eax, %esp
-
-	# ds, es = user data
 	movw $(0x20 | 3), %bx
 	movw %bx, %ds
 	movw %bx, %es
-	# gs = thread local
 	movw $(0x30 | 3), %bx
-	movw %bx, %gs
-	# fs = 0
-	xorw %bx, %bx
 	movw %bx, %fs
+	movw $(0x38 | 3), %bx
+	movw %bx, %gs
 
 	iret

kernel/arch/i686/User.S

@@ -0,0 +1,54 @@
+# bool safe_user_memcpy(void*, const void*, size_t)
+.global safe_user_memcpy
+.global safe_user_memcpy_end
+.global safe_user_memcpy_fault
+safe_user_memcpy:
+	xorl %eax, %eax
+	xchgl 4(%esp), %edi
+	xchgl 8(%esp), %esi
+	movl 12(%esp), %ecx
+	movl %edi, %edx
+	rep movsb
+	movl 4(%esp), %edi
+	movl 8(%esp), %esi
+	incl %eax
+safe_user_memcpy_fault:
+	ret
+safe_user_memcpy_end:
+
+# bool safe_user_strncpy(void*, const void*, size_t)
+.global safe_user_strncpy
+.global safe_user_strncpy_end
+.global safe_user_strncpy_fault
+safe_user_strncpy:
+	xchgl 4(%esp), %edi
+	xchgl 8(%esp), %esi
+	movl 12(%esp), %ecx
+
+	testl %ecx, %ecx
+	jz safe_user_strncpy_fault
+
+ .safe_user_strncpy_loop:
+	movb (%esi), %al
+	movb %al, (%edi)
+	testb %al, %al
+	jz .safe_user_strncpy_done
+
+	incl %edi
+	incl %esi
+	decl %ecx
+	jnz .safe_user_strncpy_loop
+
+ safe_user_strncpy_fault:
+	xorl %eax, %eax
+	jmp .safe_user_strncpy_return
+
+ .safe_user_strncpy_done:
+	movl $1, %eax
+
+ .safe_user_strncpy_return:
+	movl 4(%esp), %edi
+	movl 8(%esp), %esi
+	ret
+
+safe_user_strncpy_end:

kernel/arch/i686/Yield.S

@@ -0,0 +1,25 @@
+.global asm_yield_trampoline
+asm_yield_trampoline:
+	leal 4(%esp), %ecx
+	movl 4(%esp), %esp
+
+	pushl -4(%ecx)
+	pushl %ecx
+	pushl %eax
+	pushl %ebx
+	pushl %esi
+	pushl %edi
+	pushl %ebp
+
+	pushl %esp
+	call scheduler_on_yield
+	addl $4, %esp
+
+	popl %ebp
+	popl %edi
+	popl %esi
+	popl %ebx
+	popl %eax
+	movl 4(%esp), %ecx
+	movl 0(%esp), %esp
+	jmp *%ecx

kernel/arch/i686/boot.S

@@ -11,9 +11,28 @@
 
 .code32
 
-# multiboot2 header
+// video mode info, page align modules
+.set multiboot_flags, (1 << 2) | (1 << 0)
+
 .section .multiboot, "aw"
-	.align 8
+multiboot_start:
+	.long 0x1BADB002
+	.long multiboot_flags
+	.long -(0x1BADB002 + multiboot_flags)
+
+	.long 0
+	.long 0
+	.long 0
+	.long 0
+	.long 0
+
+	.long 0
+	.long FB_WIDTH
+	.long FB_HEIGHT
+	.long FB_BPP
+multiboot_end:
+
+.section .multiboot2, "aw"
 multiboot2_start:
 	.long 0xE85250D6
 	.long 0
@@ -49,7 +68,6 @@ multiboot2_start:
 multiboot2_end:
 
 .section .bananboot, "aw"
-	.align 8
 bananboot_start:
 	.long 0xBABAB007
 	.long -(0xBABAB007 + FB_WIDTH + FB_HEIGHT + FB_BPP)
@@ -167,6 +185,13 @@ enable_sse:
 	movl %eax, %cr4
 	ret
 
+enable_tsc:
+	# allow userspace to use RDTSC
+	movl %cr4, %ecx
+	andl $0xFFFFFFFB, %ecx
+	movl %ecx, %cr4
+	ret
+
 initialize_paging:
 	# enable PAE
 	movl %cr4, %ecx
@@ -209,6 +234,7 @@ gdt_flush:
 	# do processor initialization
 	call check_requirements
 	call enable_sse
+	call enable_tsc
 	call initialize_paging
 
 	# load higher half stack pointer
@@ -285,6 +311,7 @@ ap_protected_mode:
 	movb $1, AP_V2P(ap_stack_loaded)
 
 	leal V2P(enable_sse),        %ecx; call *%ecx
+	leal V2P(enable_tsc),        %ecx; call *%ecx
 	leal V2P(initialize_paging), %ecx; call *%ecx
 
 	# load boot gdt and enter long mode

kernel/arch/i686/interrupts.S

@@ -1,66 +1,67 @@
-.macro push_userspace
+.macro intr_header, n
-	pushw %gs
-	pushw %fs
-	pushw %es
-	pushw %ds
 	pushal
-.endm
+	testb $3, \n+8*4(%esp)
-
+	jz 1f
-.macro load_kernel_segments
 	movw $0x10, %ax
 	movw %ax, %ds
 	movw %ax, %es
 	movw %ax, %fs
-
 	movw $0x28, %ax
 	movw %ax, %gs
+1:	cld
 .endm
 
-.macro pop_userspace
+.macro intr_footer, n
-	popal
+	testb $3, \n+8*4(%esp)
-	popw %ds
+	jz 1f
-	popw %es
+	call cpp_check_signal
-	popw %fs
+	movw $(0x20 | 3), %bx
-	popw %gs
+	movw %bx, %ds
+	movw %bx, %es
+	movw $(0x30 | 3), %bx
+	movw %bx, %fs
+	movw $(0x38 | 3), %bx
+	movw %bx, %gs
+1:	popal
 .endm
 
 isr_stub:
-	push_userspace
+	intr_header 12
-	load_kernel_segments
-	cld
-
 	movl %cr0, %eax; pushl %eax
 	movl %cr2, %eax; pushl %eax
 	movl %cr3, %eax; pushl %eax
 	movl %cr4, %eax; pushl %eax
 
-	movl %esp,     %eax // register ptr
+	movl 48(%esp), %edi // isr number
-	leal 64(%esp), %ebx // interrupt stack ptr
+	movl 52(%esp), %esi // error code
-	movl 60(%esp), %ecx // error code
+	leal 56(%esp), %edx // interrupt stack ptr
-	movl 56(%esp), %edx // isr number
+	movl %esp,     %ecx // register ptr
+
+	# stack frame for stack trace
+	leal 56(%esp), %eax
+	pushl (%eax)
+	pushl %ebp
 
 	movl %esp, %ebp
 	andl $-16, %esp
 
-	pushl %eax
-	pushl %ebx
 	pushl %ecx
 	pushl %edx
+	pushl %esi
+	pushl %edi
 	call cpp_isr_handler
 
 	movl %ebp, %esp
-	addl $16, %esp
+	addl $24, %esp
 
-	pop_userspace
+	intr_footer 12
 	addl $8, %esp
 	iret
 
 irq_stub:
-	push_userspace
+	intr_header 12
-	load_kernel_segments
-	cld
 
-	movl 40(%esp), %edi # interrupt number
+	movl 32(%esp), %edi # interrupt number
 
 	movl %esp, %ebp
 	andl $-16, %esp
@@ -71,37 +72,13 @@ irq_stub:
 
 	movl %ebp, %esp
 
-	pop_userspace
+	intr_footer 12
 	addl $8, %esp
 	iret
 
-.global asm_yield_handler
-asm_yield_handler:
-	# This can only be called from kernel, so no segment saving is needed
-	pushal
-	cld
-
-	leal 32(%esp), %edi # interrupt stack ptr
-	movl %esp,     %esi # interrupt registers ptr
-
-	movl %esp, %ebp
-	andl $-16, %esp
-
-	subl $8, %esp
-	pushl %esi
-	pushl %edi
-	call cpp_yield_handler
-
-	movl %ebp, %esp
-
-	popal
-	iret
-
 .global asm_ipi_handler
 asm_ipi_handler:
-	push_userspace
+	intr_header 4
-	load_kernel_segments
-	cld
 
 	movl %esp, %ebp
 	andl $-16, %esp
@@ -110,14 +87,12 @@ asm_ipi_handler:
 
 	movl %ebp, %esp
 
-	pop_userspace
+	intr_footer 4
 	iret
 
 .global asm_timer_handler
 asm_timer_handler:
-	push_userspace
+	intr_header 4
-	load_kernel_segments
-	cld
 
 	movl %esp, %ebp
 	andl $-16, %esp
@@ -126,7 +101,7 @@ asm_timer_handler:
 
 	movl %ebp, %esp
 
-	pop_userspace
+	intr_footer 4
 	iret
 
 .macro isr n

kernel/arch/i686/linker.ld

@@ -11,6 +11,7 @@ SECTIONS
 	{
 		g_kernel_execute_start = .;
 		*(.multiboot)
+		*(.multiboot2)
 		*(.bananboot)
 		*(.text.*)
 	}

kernel/arch/x86_64/PageTable.cpp

@@ -23,7 +23,7 @@ namespace Kernel
 	SpinLock PageTable::s_fast_page_lock;
 
 	static constexpr vaddr_t s_hhdm_offset = 0xFFFF800000000000;
-	static bool s_is_hddm_initialized = false;
+	static bool s_is_post_heap_done = false;
 
 	constexpr uint64_t s_page_flag_mask = 0x8000000000000FFF;
 	constexpr uint64_t s_page_addr_mask = ~s_page_flag_mask;
@@ -376,7 +376,7 @@ namespace Kernel
 		V2P                               = &FuncsHHDM::V2P;
 		P2V                               = &FuncsHHDM::P2V;
 
-		s_is_hddm_initialized = true;
+		s_is_post_heap_done = true;
 
 		// This is a hack to unmap fast page. fast page pt is copied
 		// while it is mapped, so we need to manually unmap it
@@ -485,6 +485,7 @@ namespace Kernel
 		constexpr uint64_t pml4e = (uc_vaddr >> 39) & 0x1FF;
 		constexpr uint64_t pdpte = (uc_vaddr >> 30) & 0x1FF;
 		constexpr uint64_t pde   = (uc_vaddr >> 21) & 0x1FF;
+		constexpr uint64_t pte   = (uc_vaddr >> 12) & 0x1FF;
 
 		uint64_t* pml4 = P2V(m_highest_paging_struct);
 		ASSERT(!(pml4[pml4e] & Flags::Present));
@@ -497,6 +498,10 @@ namespace Kernel
 		uint64_t* pd = P2V(pdpt[pdpte] & s_page_addr_mask);
 		ASSERT(!(pd[pde] & Flags::Present));
 		pd[pde] = allocate_zeroed_page_aligned_page() | Flags::ReadWrite | Flags::Present;
+
+		uint64_t* pt = P2V(pd[pde] & s_page_addr_mask);
+		ASSERT(pt[pte] == 0);
+		pt[pte] = Flags::Reserved;
 	}
 
 	void PageTable::map_fast_page(paddr_t paddr)
@@ -521,7 +526,7 @@ namespace Kernel
 		ASSERT(!(pt[pte] & Flags::Present));
 		pt[pte] = paddr | Flags::ReadWrite | Flags::Present;
 
-		invalidate(fast_page(), false);
+		asm volatile("invlpg (%0)" :: "r"(fast_page()) : "memory");
 	}
 
 	void PageTable::unmap_fast_page()
@@ -542,9 +547,9 @@ namespace Kernel
 		      uint64_t* pt   = P2V(pd[pde] & s_page_addr_mask);
 
 		ASSERT(pt[pte] & Flags::Present);
-		pt[pte] = 0;
+		pt[pte] = Flags::Reserved;
 
-		invalidate(fast_page(), false);
+		asm volatile("invlpg (%0)" :: "r"(fast_page()) : "memory");
 	}
 
 	BAN::ErrorOr<PageTable*> PageTable::create_userspace()
@@ -612,10 +617,39 @@ namespace Kernel
 		Processor::set_current_page_table(this);
 	}
 
-	void PageTable::invalidate(vaddr_t vaddr, bool send_smp_message)
+	void PageTable::invalidate_range(vaddr_t vaddr, size_t pages, bool send_smp_message)
 	{
 		ASSERT(vaddr % PAGE_SIZE == 0);
-		asm volatile("invlpg (%0)" :: "r"(vaddr) : "memory");
+
+		const bool is_userspace = (vaddr < KERNEL_OFFSET);
+		if (is_userspace && this != &PageTable::current())
+			;
+		else if (pages <= 32 || !s_is_post_heap_done)
+		{
+			for (size_t i = 0; i < pages; i++, vaddr += PAGE_SIZE)
+				asm volatile("invlpg (%0)" :: "r"(vaddr));
+		}
+		else if (is_userspace || !s_has_pge)
+		{
+			asm volatile("movq %0, %%cr3" :: "r"(m_highest_paging_struct));
+		}
+		else
+		{
+			asm volatile(
+				"movq %%cr4, %%rax;"
+
+				"andq $~0x80, %%rax;"
+				"movq %%rax, %%cr4;"
+
+				"movq %0, %%cr3;"
+
+				"orq $0x80, %%rax;"
+				"movq %%rax, %%cr4;"
+				:
+				: "r"(m_highest_paging_struct)
+				: "rax"
+			);
+		}
 
 		if (send_smp_message)
 		{
@@ -623,13 +657,14 @@ namespace Kernel
 				.type = Processor::SMPMessage::Type::FlushTLB,
 				.flush_tlb = {
 					.vaddr      = vaddr,
-					.page_count = 1
+					.page_count = pages,
+					.page_table = vaddr < KERNEL_OFFSET ? this : nullptr,
 				}
 			});
 		}
 	}
 
-	void PageTable::unmap_page(vaddr_t vaddr, bool send_smp_message)
+	void PageTable::unmap_page(vaddr_t vaddr, bool invalidate)
 	{
 		ASSERT(vaddr);
 		ASSERT(vaddr != fast_page());
@@ -658,30 +693,27 @@ namespace Kernel
 		uint64_t* pd   = P2V(pdpt[pdpte] & s_page_addr_mask);
 		uint64_t* pt   = P2V(pd[pde]     & s_page_addr_mask);
 
+		const paddr_t old_paddr = pt[pte] & PAGE_ADDR_MASK;
+
 		pt[pte] = 0;
-		invalidate(vaddr, send_smp_message);
+
+		if (invalidate && old_paddr != 0)
+			invalidate_page(vaddr, true);
 	}
 
 	void PageTable::unmap_range(vaddr_t vaddr, size_t size)
 	{
 		ASSERT(vaddr % PAGE_SIZE == 0);
 
-		size_t page_count = range_page_count(vaddr, size);
+		const size_t page_count = range_page_count(vaddr, size);
 
 		SpinLockGuard _(m_lock);
 		for (vaddr_t page = 0; page < page_count; page++)
 			unmap_page(vaddr + page * PAGE_SIZE, false);
-
+		invalidate_range(vaddr, page_count, true);
-		Processor::broadcast_smp_message({
-			.type = Processor::SMPMessage::Type::FlushTLB,
-			.flush_tlb = {
-				.vaddr      = vaddr,
-				.page_count = page_count
-			}
-		});
 	}
 
-	void PageTable::map_page_at(paddr_t paddr, vaddr_t vaddr, flags_t flags, MemoryType memory_type, bool send_smp_message)
+	void PageTable::map_page_at(paddr_t paddr, vaddr_t vaddr, flags_t flags, MemoryType memory_type, bool invalidate)
 	{
 		ASSERT(vaddr);
 		ASSERT(vaddr != fast_page());
@@ -742,9 +774,12 @@ namespace Kernel
 		if (!(flags & Flags::Present))
 			uwr_flags &= ~Flags::Present;
 
+		const paddr_t old_paddr = pt[pte] & PAGE_ADDR_MASK;
+
 		pt[pte] = paddr | uwr_flags | extra_flags;
 
-		invalidate(vaddr, send_smp_message);
+		if (invalidate && old_paddr != 0)
+			invalidate_page(vaddr, true);
 	}
 
 	void PageTable::map_range_at(paddr_t paddr, vaddr_t vaddr, size_t size, flags_t flags, MemoryType memory_type)
@@ -760,14 +795,66 @@ namespace Kernel
 		SpinLockGuard _(m_lock);
 		for (size_t page = 0; page < page_count; page++)
 			map_page_at(paddr + page * PAGE_SIZE, vaddr + page * PAGE_SIZE, flags, memory_type, false);
+		invalidate_range(vaddr, page_count, true);
+	}
 
-		Processor::broadcast_smp_message({
+	void PageTable::remove_writable_from_range(vaddr_t vaddr, size_t size)
-			.type = Processor::SMPMessage::Type::FlushTLB,
+	{
-			.flush_tlb = {
+		ASSERT(vaddr);
-				.vaddr      = vaddr,
+		ASSERT(vaddr % PAGE_SIZE == 0);
-				.page_count = page_count
+
+		ASSERT(is_canonical(vaddr));
+		ASSERT(is_canonical(vaddr + size - 1));
+
+		const vaddr_t uc_vaddr_start = uncanonicalize(vaddr);
+		const vaddr_t uc_vaddr_end   = uncanonicalize(vaddr + size - 1);
+
+		uint16_t pml4e = (uc_vaddr_start >> 39) & 0x1FF;
+		uint16_t pdpte = (uc_vaddr_start >> 30) & 0x1FF;
+		uint16_t pde   = (uc_vaddr_start >> 21) & 0x1FF;
+		uint16_t pte   = (uc_vaddr_start >> 12) & 0x1FF;
+
+		const uint16_t e_pml4e = (uc_vaddr_end >> 39) & 0x1FF;
+		const uint16_t e_pdpte = (uc_vaddr_end >> 30) & 0x1FF;
+		const uint16_t e_pde   = (uc_vaddr_end >> 21) & 0x1FF;
+		const uint16_t e_pte   = (uc_vaddr_end >> 12) & 0x1FF;
+
+		SpinLockGuard _(m_lock);
+
+		const uint64_t* pml4 = P2V(m_highest_paging_struct);
+		for (; pml4e <= e_pml4e; pml4e++)
+		{
+			if (!(pml4[pml4e] & Flags::ReadWrite))
+				continue;
+			const uint64_t* pdpt = P2V(pml4[pml4e] & s_page_addr_mask);
+			for (; pdpte < 512; pdpte++)
+			{
+				if (pml4e == e_pml4e && pdpte > e_pdpte)
+					break;
+				if (!(pdpt[pdpte] & Flags::ReadWrite))
+					continue;
+				const uint64_t* pd = P2V(pdpt[pdpte] & s_page_addr_mask);
+				for (; pde < 512; pde++)
+				{
+					if (pml4e == e_pml4e && pdpte == e_pdpte && pde > e_pde)
+						break;
+					if (!(pd[pde] & Flags::ReadWrite))
+						continue;
+					uint64_t* pt = P2V(pd[pde] & s_page_addr_mask);
+					for (; pte < 512; pte++)
+					{
+						if (pml4e == e_pml4e && pdpte == e_pdpte && pde == e_pde && pte > e_pte)
+							break;
+						pt[pte] &= ~static_cast<uint64_t>(Flags::ReadWrite);
+					}
+					pte = 0;
+				}
+				pde = 0;
 			}
-		});
+			pdpte = 0;
+		}
+
+		invalidate_range(vaddr, size / PAGE_SIZE, true);
 	}
 
 	uint64_t PageTable::get_page_data(vaddr_t vaddr) const
@@ -814,13 +901,13 @@ namespace Kernel
 		return page_data & s_page_addr_mask;
 	}
 
-	bool PageTable::reserve_page(vaddr_t vaddr, bool only_free)
+	bool PageTable::reserve_page(vaddr_t vaddr, bool only_free, bool invalidate)
 	{
 		SpinLockGuard _(m_lock);
 		ASSERT(vaddr % PAGE_SIZE == 0);
 		if (only_free && !is_page_free(vaddr))
 			return false;
-		map_page_at(0, vaddr, Flags::Reserved);
+		map_page_at(0, vaddr, Flags::Reserved, MemoryType::Normal, invalidate);
 		return true;
 	}
 
@@ -834,7 +921,8 @@ namespace Kernel
 		if (only_free && !is_range_free(vaddr, bytes))
 			return false;
 		for (size_t offset = 0; offset < bytes; offset += PAGE_SIZE)
-			reserve_page(vaddr + offset);
+			reserve_page(vaddr + offset, true, false);
+		invalidate_range(vaddr, bytes / PAGE_SIZE, true);
 		return true;
 	}
 
@@ -848,9 +936,9 @@ namespace Kernel
 			last_address -= rem;
 
 		ASSERT(is_canonical(first_address));
-		ASSERT(is_canonical(last_address));
+		ASSERT(is_canonical(last_address - 1));
 		const vaddr_t uc_vaddr_start = uncanonicalize(first_address);
-		const vaddr_t uc_vaddr_end   = uncanonicalize(last_address);
+		const vaddr_t uc_vaddr_end   = uncanonicalize(last_address - 1);
 
 		uint16_t pml4e = (uc_vaddr_start >> 39) & 0x1FF;
 		uint16_t pdpte = (uc_vaddr_start >> 30) & 0x1FF;
@@ -867,10 +955,8 @@ namespace Kernel
 		// Try to find free page that can be mapped without
 		// allocations (page table with unused entries)
 		const uint64_t* pml4 = P2V(m_highest_paging_struct);
-		for (; pml4e < 512; pml4e++)
+		for (; pml4e <= e_pml4e; pml4e++)
 		{
-			if (pml4e > e_pml4e)
-				break;
 			if (!(pml4[pml4e] & Flags::Present))
 				continue;
 			const uint64_t* pdpt = P2V(pml4[pml4e] & s_page_addr_mask);
@@ -890,22 +976,24 @@ namespace Kernel
 					const uint64_t* pt = P2V(pd[pde] & s_page_addr_mask);
 					for (; pte < 512; pte++)
 					{
-						if (pml4e == e_pml4e && pdpte == e_pdpte && pde == e_pde && pte >= e_pte)
+						if (pml4e == e_pml4e && pdpte == e_pdpte && pde == e_pde && pte > e_pte)
 							break;
-						if (!(pt[pte] & Flags::Used))
+						if (pt[pte] & Flags::Used)
-						{
+							continue;
-							vaddr_t vaddr = 0;
+						vaddr_t vaddr = 0;
-							vaddr |= static_cast<uint64_t>(pml4e) << 39;
+						vaddr |= static_cast<uint64_t>(pml4e) << 39;
-							vaddr |= static_cast<uint64_t>(pdpte) << 30;
+						vaddr |= static_cast<uint64_t>(pdpte) << 30;
-							vaddr |= static_cast<uint64_t>(pde)   << 21;
+						vaddr |= static_cast<uint64_t>(pde)   << 21;
-							vaddr |= static_cast<uint64_t>(pte)   << 12;
+						vaddr |= static_cast<uint64_t>(pte)   << 12;
-							vaddr = canonicalize(vaddr);
+						vaddr = canonicalize(vaddr);
-							ASSERT(reserve_page(vaddr));
+						ASSERT(reserve_page(vaddr));
-							return vaddr;
+						return vaddr;
-						}
 					}
+					pte = 0;
 				}
+				pde = 0;
 			}
+			pdpte = 0;
 		}
 
 		for (vaddr_t uc_vaddr = uc_vaddr_start; uc_vaddr < uc_vaddr_end; uc_vaddr += PAGE_SIZE)
@@ -917,7 +1005,7 @@ namespace Kernel
 			}
 		}
 
-		ASSERT_NOT_REACHED();
+		return 0;
 	}
 
 	vaddr_t PageTable::reserve_free_contiguous_pages(size_t page_count, vaddr_t first_address, vaddr_t last_address)
@@ -930,7 +1018,7 @@ namespace Kernel
 			last_address -= rem;
 
 		ASSERT(is_canonical(first_address));
-		ASSERT(is_canonical(last_address));
+		ASSERT(is_canonical(last_address - 1));
 
 		SpinLockGuard _(m_lock);
 
@@ -959,7 +1047,7 @@ namespace Kernel
 			}
 		}
 
-		ASSERT_NOT_REACHED();
+		return 0;
 	}
 
 	bool PageTable::is_page_free(vaddr_t page) const

kernel/arch/x86_64/Signal.S

@@ -1,30 +1,48 @@
 .section .userspace, "ax"
 
 // stack contains
-//    return address
+//    (8 bytes)  return address (on return stack)
-//    signal number
+//    (8 bytes)  return stack
-//    signal handler
+//    (8 bytes)  return rflags
+//    (8 bytes)  restore sigmask
+//    (56 bytes) siginfo_t
+//    (8 bytes)  signal number
+//    (8 bytes)  signal handler
 
 .global signal_trampoline
 signal_trampoline:
-	pushq %rax
+	pushq %r15 // gregs
-	pushq %rbx
-	pushq %rcx
-	pushq %rdx
-	pushq %rbp
-	pushq %rdi
-	pushq %rsi
-	pushq %r8
-	pushq %r9
-	pushq %r10
-	pushq %r11
-	pushq %r12
-	pushq %r13
 	pushq %r14
-	pushq %r15
+	pushq %r13
+	pushq %r12
+	pushq %r11
+	pushq %r10
+	pushq %r9
+	pushq %r8
+	pushq %rsi
+	pushq %rdi
+	pushq %rdx
+	pushq %rcx
+	pushq %rbx
+	pushq %rax
+	pushq %rbp
 
-	movq 128(%rsp), %rdi
+	movq 208(%rsp), %rax
-	movq 120(%rsp), %rax
+	pushq %rax; addq $(128 + 8), (%rsp)
+	pushq (%rax)
+
+	// FIXME: populate these
+	xorq %rax, %rax
+	pushq %rax // stack
+	pushq %rax
+	pushq %rax
+	pushq %rax // sigset
+	pushq %rax // link
+
+	movq  %rsp,     %rdx // ucontext
+	leaq 192(%rsp), %rsi // siginfo
+	movq 184(%rsp), %rdi // signal number
+	movq 176(%rsp), %rax // handler
 
 	// align stack to 16 bytes
 	movq %rsp, %rbp
@@ -40,24 +58,40 @@ signal_trampoline:
 
 	// restore stack
 	movq %rbp, %rsp
-	popq %r15
+	addq $40, %rsp
-	popq %r14
-	popq %r13
-	popq %r12
-	popq %r11
-	popq %r10
-	popq %r9
-	popq %r8
-	popq %rsi
-	popq %rdi
-	popq %rbp
-	popq %rdx
-	popq %rcx
-	popq %rbx
-	popq %rax
 
+	// restore sigmask
+	movq $83, %rdi       // SYS_SIGPROCMASK
+	movq $3,  %rsi       // SIG_SETMASK
+	leaq 192(%rsp), %rdx // set
+	xorq %r10, %r10      // oset
+	syscall
+
+	// restore registers
 	addq $16, %rsp
+	popq %rbp
+	popq %rax
+	popq %rbx
+	popq %rcx
+	popq %rdx
+	popq %rdi
+	popq %rsi
+	popq %r8
+	popq %r9
+	popq %r10
+	popq %r11
+	popq %r12
+	popq %r13
+	popq %r14
+	popq %r15
+
+	// skip handler, number, siginfo_t, sigmask
+	addq $80, %rsp
+
+	// restore flags
 	popfq
 
+	movq (%rsp), %rsp
+
 	// return over red-zone
 	ret $128

kernel/arch/x86_64/Syscall.S

@@ -1,49 +1,26 @@
-// arguments in RAX, RBX, RCX, RDX, RSI, RDI
-// System V ABI: RDI, RSI, RDX, RCX, R8, R9
 .global asm_syscall_handler
 asm_syscall_handler:
-	pushq %rbx
+	swapgs
-	pushq %rcx
+
-	pushq %rdx
+	movq %rsp, %rax
-	pushq %rdi
+	movq %gs:8, %rsp
-	pushq %rsi
+
-	pushq %rbp
+	pushq $(0x20 | 3)
-	pushq %r8
+	pushq %rax
-	pushq %r9
-	pushq %r10
 	pushq %r11
-	pushq %r12
+	pushq $(0x28 | 3)
-	pushq %r13
+	pushq %rcx
-	pushq %r14
+	subq $8, %rsp
-	pushq %r15
-	cld
 
-	movq %rsi, %r8
+	movq %r10, %rcx
-	movq %rdi, %r9
-	movq %rax, %rdi
-	movq %rbx, %rsi
-	xchgq %rcx, %rdx
-	leaq 112(%rsp), %rbx
-	pushq %rbx
 	call cpp_syscall_handler
-	addq $8, %rsp
 
-	popq %r15
+	movq  8(%rsp), %rcx
-	popq %r14
+	movq 24(%rsp), %r11
-	popq %r13
+	movq 32(%rsp), %rsp
-	popq %r12
-	popq %r11
-	popq %r10
-	popq %r9
-	popq %r8
-	popq %rbp
-	popq %rsi
-	popq %rdi
-	popq %rdx
-	popq %rcx
-	popq %rbx
-	iretq
 
+	swapgs
+	sysretq
 
 .global sys_fork_trampoline
 sys_fork_trampoline:
@@ -56,7 +33,7 @@ sys_fork_trampoline:
 
 	call read_ip
 	testq %rax, %rax
-	je .reload_stack
+	jz .done
 
 	movq %rax, %rsi
 	movq %rsp, %rdi
@@ -70,9 +47,3 @@ sys_fork_trampoline:
 	popq %rbp
 	popq %rbx
 	ret
-
- .reload_stack:
-	call get_thread_start_sp
-	movq %rax, %rsp
-	xorq %rax, %rax
-	jmp .done

kernel/arch/x86_64/Thread.S

@@ -7,9 +7,6 @@ read_ip:
 # void start_kernel_thread()
 .global start_kernel_thread
 start_kernel_thread:
-	call get_thread_start_sp
-	movq %rax, %rsp
-
 	# STACK LAYOUT
 	#   on_exit arg
 	#   on_exit func
@@ -27,9 +24,5 @@ start_kernel_thread:
 
 .global start_userspace_thread
 start_userspace_thread:
-	call load_thread_sse
+	swapgs
-
-	call get_thread_start_sp
-	movq %rax, %rsp
-
 	iretq

kernel/arch/x86_64/User.S

@@ -0,0 +1,87 @@
+# bool safe_user_memcpy(void*, const void*, size_t)
+.global safe_user_memcpy
+.global safe_user_memcpy_end
+.global safe_user_memcpy_fault
+safe_user_memcpy:
+	xorq %rax, %rax
+	movq %rdx, %rcx
+	rep movsb
+	incq %rax
+safe_user_memcpy_fault:
+	ret
+safe_user_memcpy_end:
+
+# bool safe_user_strncpy(void*, const void*, size_t)
+.global safe_user_strncpy
+.global safe_user_strncpy_end
+.global safe_user_strncpy_fault
+safe_user_strncpy:
+	movq %rdx, %rcx
+	testq %rcx, %rcx
+	jz safe_user_strncpy_fault
+
+ .safe_user_strncpy_align_loop:
+	testb $0x7, %sil
+	jz .safe_user_strncpy_align_done
+
+	movb (%rsi), %al
+	movb %al, (%rdi)
+	testb %al, %al
+	jz .safe_user_strncpy_done
+
+	incq %rdi
+	incq %rsi
+	decq %rcx
+	jnz .safe_user_strncpy_align_loop
+	jmp safe_user_strncpy_fault
+
+ .safe_user_strncpy_align_done:
+	movq $0x0101010101010101, %r8
+	movq $0x8080808080808080, %r9
+
+ .safe_user_strncpy_qword_loop:
+	cmpq $8, %rcx
+	jb .safe_user_strncpy_qword_done
+
+	movq (%rsi), %rax
+	movq %rax, %r10
+	movq %rax, %r11
+
+	# https://graphics.stanford.edu/~seander/bithacks.html#ZeroInWord
+	subq %r8, %r10
+	notq %r11
+	andq %r11, %r10
+	andq %r9, %r10
+	jnz .safe_user_strncpy_byte_loop
+
+	movq %rax, (%rdi)
+
+	addq $8, %rdi
+	addq $8, %rsi
+	subq $8, %rcx
+	jnz .safe_user_strncpy_qword_loop
+	jmp safe_user_strncpy_fault
+
+ .safe_user_strncpy_qword_done:
+	testq %rcx, %rcx
+	jz safe_user_strncpy_fault
+
+ .safe_user_strncpy_byte_loop:
+	movb (%rsi), %al
+	movb %al, (%rdi)
+	testb %al, %al
+	jz .safe_user_strncpy_done
+
+	incq %rdi
+	incq %rsi
+	decq %rcx
+	jnz .safe_user_strncpy_byte_loop
+
+ safe_user_strncpy_fault:
+	xorq %rax, %rax
+	ret
+
+ .safe_user_strncpy_done:
+	movb $1, %al
+	ret
+safe_user_strncpy_end:

kernel/arch/x86_64/Yield.S

@@ -0,0 +1,29 @@
+.global asm_yield_trampoline
+asm_yield_trampoline:
+	leaq 8(%rsp), %rcx
+	movq %rdi, %rsp
+
+	subq $8, %rsp
+	pushq -8(%rcx)
+	pushq %rcx
+	pushq %rax
+	pushq %rbx
+	pushq %rbp
+	pushq %r12
+	pushq %r13
+	pushq %r14
+	pushq %r15
+
+	movq %rsp, %rdi
+	call scheduler_on_yield
+
+	popq %r15
+	popq %r14
+	popq %r13
+	popq %r12
+	popq %rbp
+	popq %rbx
+	popq %rax
+	movq 8(%rsp), %rcx
+	movq 0(%rsp), %rsp
+	jmp *%rcx

kernel/arch/x86_64/boot.S

@@ -11,9 +11,28 @@
 
 .code32
 
-# multiboot2 header
+// custom addresses, video mode info, page align modules
+.set multiboot_flags, (1 << 16) | (1 << 2) | (1 << 0)
+
 .section .multiboot, "aw"
-	.align 8
+multiboot_start:
+	.long 0x1BADB002
+	.long multiboot_flags
+	.long -(0x1BADB002 + multiboot_flags)
+
+	.long V2P(multiboot_start)
+	.long V2P(g_kernel_start)
+	.long V2P(g_kernel_bss_start)
+	.long V2P(g_kernel_end)
+	.long V2P(_start)
+
+	.long 0
+	.long FB_WIDTH
+	.long FB_HEIGHT
+	.long FB_BPP
+multiboot_end:
+
+.section .multiboot2, "aw"
 multiboot2_start:
 	.long 0xE85250D6
 	.long 0
@@ -49,7 +68,6 @@ multiboot2_start:
 multiboot2_end:
 
 .section .bananboot, "aw"
-	.align 8
 bananboot_start:
 	.long 0xBABAB007
 	.long -(0xBABAB007 + FB_WIDTH + FB_HEIGHT + FB_BPP)
@@ -162,6 +180,13 @@ enable_sse:
 	movl %eax, %cr4
 	ret
 
+enable_tsc:
+	# allow userspace to use RDTSC
+	movl %cr4, %ecx
+	andl $0xFFFFFFFB, %ecx
+	movl %ecx, %cr4
+	ret
+
 initialize_paging:
 	# enable PAE
 	movl %cr4, %ecx
@@ -198,6 +223,7 @@ _start:
 
 	call check_requirements
 	call enable_sse
+	call enable_tsc
 	call initialize_paging
 
 	# flush gdt and jump to 64 bit
@@ -284,6 +310,7 @@ ap_protected_mode:
 	movb $1, AP_V2P(ap_stack_loaded)
 
 	leal V2P(enable_sse),        %ecx; call *%ecx
+	leal V2P(enable_tsc),        %ecx; call *%ecx
 	leal V2P(initialize_paging), %ecx; call *%ecx
 
 	# load boot gdt and enter long mode

kernel/arch/x86_64/interrupts.S

@@ -1,4 +1,4 @@
-.macro pushaq
+.macro intr_header, n
 	pushq %rax
 	pushq %rcx
 	pushq %rdx
@@ -14,10 +14,18 @@
 	pushq %r13
 	pushq %r14
 	pushq %r15
+	testb $3, \n+15*8(%rsp)
+	jz 1f
+	swapgs
+1:	cld
 .endm
 
-.macro popaq
+.macro intr_footer, n
-	popq %r15
+	testb $3, \n+15*8(%rsp)
+	jz 1f
+	call cpp_check_signal
+	swapgs
+1:	popq %r15
 	popq %r14
 	popq %r13
 	popq %r12
@@ -35,8 +43,7 @@
 .endm
 
 isr_stub:
-	pushaq
+	intr_header 24
-	cld
 	movq %cr0, %rax; pushq %rax
 	movq %cr2, %rax; pushq %rax
 	movq %cr3, %rax; pushq %rax
@@ -49,43 +56,33 @@ isr_stub:
 	call cpp_isr_handler
 	addq $32, %rsp
 
-	popaq
+	intr_footer 24
 	addq $16, %rsp
 	iretq
 
 irq_stub:
-	pushaq
+	intr_header 24
-	cld
+	xorq %rbp, %rbp
 	movq 120(%rsp), %rdi	# irq number
 	call cpp_irq_handler
-	popaq
+	intr_footer 24
 	addq $16, %rsp
 	iretq
 
-.global asm_yield_handler
-asm_yield_handler:
-	pushaq
-	cld
-	leaq 120(%rsp), %rdi # interrupt stack ptr
-	movq     %rsp,  %rsi # interrupt register ptr
-	call cpp_yield_handler
-	popaq
-	iretq
-
 .global asm_ipi_handler
 asm_ipi_handler:
-	pushaq
+	intr_header 8
-	cld
+	xorq %rbp, %rbp
 	call cpp_ipi_handler
-	popaq
+	intr_footer 8
 	iretq
 
 .global asm_timer_handler
 asm_timer_handler:
-	pushaq
+	intr_header 8
-	cld
+	xorq %rbp, %rbp
 	call cpp_timer_handler
-	popaq
+	intr_footer 8
 	iretq
 
 .macro isr n

kernel/arch/x86_64/linker.ld

@@ -11,6 +11,7 @@ SECTIONS
 	{
 		g_kernel_execute_start = .;
 		*(.multiboot)
+		*(.multiboot2)
 		*(.bananboot)
 		*(.text.*)
 	}
@@ -43,6 +44,7 @@ SECTIONS
 	}
 	.bss ALIGN(4K) : AT(ADDR(.bss) - KERNEL_OFFSET)
 	{
+		g_kernel_bss_start = .;
 		*(COMMON)
 		*(.bss)
 		g_kernel_writable_end = .;

kernel/include/kernel/ACPI/ACPI.h

@@ -2,6 +2,7 @@
 
 #include <BAN/Vector.h>
 #include <kernel/ACPI/AML/Namespace.h>
+#include <kernel/ACPI/EmbeddedController.h>
 #include <kernel/ACPI/Headers.h>
 #include <kernel/Memory/Types.h>
 #include <kernel/ThreadBlocker.h>
@@ -35,8 +36,12 @@ namespace Kernel::ACPI
 		BAN::ErrorOr<void> poweroff();
 		BAN::ErrorOr<void> reset();
 
+		BAN::ErrorOr<void> register_gpe_handler(uint8_t gpe, void (*callback)(void*), void* argument);
+
 		void handle_irq() override;
 
+		BAN::Span<BAN::UniqPtr<EmbeddedController>> embedded_controllers() { return m_embedded_controllers.span(); }
+
 	private:
 		ACPI() = default;
 		BAN::ErrorOr<void> initialize_impl();
@@ -50,6 +55,12 @@ namespace Kernel::ACPI
 
 		BAN::ErrorOr<void> route_interrupt_link_device(const AML::Scope& device, uint64_t& routed_irq_mask);
 
+		BAN::ErrorOr<void> initialize_embedded_controller(const AML::Scope& embedded_controller);
+		BAN::ErrorOr<void> initialize_embedded_controllers();
+
+		BAN::Optional<GAS> find_gpe_block(size_t index);
+		bool enable_gpe(uint8_t gpe);
+
 	private:
 		paddr_t m_header_table_paddr = 0;
 		vaddr_t m_header_table_vaddr = 0;
@@ -68,8 +79,23 @@ namespace Kernel::ACPI
 
 		ThreadBlocker m_event_thread_blocker;
 
+		BAN::Vector<BAN::UniqPtr<EmbeddedController>> m_embedded_controllers;
+
+		struct GPEHandler
+		{
+			bool has_callback { false };
+			union {
+				AML::Reference* method;
+				struct
+				{
+					void (*callback)(void*);
+					void* argument;
+				};
+			};
+		};
+		bool m_has_any_gpes { false };
 		AML::Scope m_gpe_scope;
-		BAN::Array<AML::Reference*, 0xFF> m_gpe_methods { nullptr };
+		BAN::Array<GPEHandler, 0xFF> m_gpe_methods;
 
 		bool m_hardware_reduced { false };
 		AML::Namespace* m_namespace { nullptr };

kernel/include/kernel/ACPI/AML/Namespace.h

@@ -50,6 +50,7 @@ namespace Kernel::ACPI::AML
 		BAN::ErrorOr<void> for_each_child(const Scope&, const BAN::Function<BAN::Iteration(const Scope&, Reference*)>&);
 
 		BAN::ErrorOr<BAN::Vector<Scope>> find_device_with_eisa_id(BAN::StringView eisa_id);
+		BAN::ErrorOr<BAN::Vector<Scope>> find_device_with_eisa_id(BAN::Span<BAN::StringView> eisa_ids);
 
 	private:
 		BAN::ErrorOr<Scope> resolve_path(const Scope& scope, const NameString& name_string);

kernel/include/kernel/ACPI/EmbeddedController.h

@@ -0,0 +1,69 @@
+#pragma once
+
+#include <BAN/Atomic.h>
+#include <BAN/UniqPtr.h>
+
+#include <kernel/ACPI/AML/Scope.h>
+#include <kernel/Lock/Mutex.h>
+#include <kernel/Thread.h>
+
+namespace Kernel::ACPI
+{
+
+	class EmbeddedController
+	{
+	public:
+		static BAN::ErrorOr<BAN::UniqPtr<EmbeddedController>> create(AML::Scope&& scope, uint16_t command_port, uint16_t data_port, BAN::Optional<uint8_t> gpe);
+		~EmbeddedController();
+
+		BAN::ErrorOr<uint8_t> read_byte(uint8_t offset);
+		BAN::ErrorOr<void> write_byte(uint8_t offset, uint8_t value);
+
+		const AML::Scope& scope() const { return m_scope; }
+
+	private:
+		EmbeddedController(AML::Scope&& scope, uint16_t command_port, uint16_t data_port, bool has_gpe)
+			: m_scope(BAN::move(scope))
+			, m_command_port(command_port)
+			, m_data_port(data_port)
+			, m_has_gpe(has_gpe)
+		{ }
+
+	private:
+		void wait_status_bit(uint8_t bit, uint8_t value);
+
+		uint8_t read_one(uint16_t port);
+		void write_one(uint16_t port, uint8_t value);
+
+		static void handle_gpe_wrapper(void*);
+		void handle_gpe();
+
+		BAN::ErrorOr<void> call_query_method(uint8_t notification);
+
+		void thread_task();
+
+		struct Command
+		{
+			uint8_t command;
+			BAN::Optional<uint8_t> data1;
+			BAN::Optional<uint8_t> data2;
+			uint8_t* response;
+			BAN::Atomic<bool> done;
+		};
+		BAN::ErrorOr<void> send_command(Command& command);
+
+	private:
+		const AML::Scope m_scope;
+		const uint16_t m_command_port;
+		const uint16_t m_data_port;
+		const bool m_has_gpe;
+
+		Mutex m_mutex;
+		ThreadBlocker m_thread_blocker;
+
+		BAN::Optional<Command*> m_queued_command;
+
+		Thread* m_thread { nullptr };
+	};
+
+}

kernel/include/kernel/ACPI/Resource.h

@@ -0,0 +1,179 @@
+#pragma once
+
+#include <BAN/ByteSpan.h>
+#include <BAN/Debug.h>
+#include <BAN/Optional.h>
+
+namespace Kernel::ACPI
+{
+
+	struct ResourceData
+	{
+		enum class Type
+		{
+			IRQ,
+			DMA,
+			IOPort,
+			FixedIOPort,
+			FixedDMA,
+
+			// TODO: large stuff the stuff :)
+		};
+
+		union {
+			struct {
+				uint16_t irq_mask;
+				union {
+					struct {
+						uint8_t edge_triggered : 1;
+						uint8_t                : 2;
+						uint8_t active_low     : 1;
+						uint8_t shared         : 1;
+						uint8_t wake_capable   : 1;
+						uint8_t                : 2;
+					};
+					uint8_t raw;
+				};
+			} irq;
+			struct {
+				uint8_t channel_mask;
+				union {
+					struct {
+						uint8_t type          : 2; // 0: 8 bit, 1: 8 and 16 bit, 2: 16 bit only
+						uint8_t bus_master    : 1;
+						uint8_t               : 2;
+						uint8_t channel_speed : 2; // 0: compatibility, 1: type A, 2: type B, 3: type F
+						uint8_t               : 1;
+					};
+					uint8_t raw;
+				};
+			} dma;
+			struct {
+				uint16_t range_min_base;
+				uint16_t range_max_base;
+				uint8_t base_alignment;
+				uint8_t range_length;
+			} io_port;
+			struct {
+				uint16_t range_base;
+				uint8_t range_length;
+			} fixed_io_port;
+			struct {
+				uint16_t request_line;
+				uint16_t channel;
+				uint8_t transfer_width; // 0: 8 bit, 1: 16 bit, 2: 32 bit, 3: 64 bit, 4: 128 bit
+			} fixed_dma;
+		} as;
+		Type type;
+	};
+
+	class ResourceParser
+	{
+	public:
+		ResourceParser(BAN::ConstByteSpan buffer)
+			: m_buffer(buffer)
+		{}
+
+		BAN::Optional<ResourceData> get_next()
+		{
+			for (;;)
+			{
+				if (m_buffer.empty())
+					return {};
+
+				if (m_buffer[0] & 0x80)
+				{
+					dprintln("Skipping large resource 0x{2H}", m_buffer[0] & 0x7F);
+					const uint16_t length = (m_buffer[2] << 8) | m_buffer[1];
+					if (m_buffer.size() < static_cast<size_t>(3 + length))
+						return {};
+					m_buffer = m_buffer.slice(3 + length);
+					continue;
+				}
+
+				const uint8_t length = m_buffer[0] & 0x07;
+				if (m_buffer.size() < static_cast<size_t>(1 + length))
+					return {};
+
+				BAN::Optional<ResourceData> result;
+				switch ((m_buffer[0] >> 3) & 0x0F)
+				{
+					case 0x04:
+						if (length < 2)
+							break;
+						result = ResourceData {
+							.as = { .irq = {
+								.irq_mask = static_cast<uint16_t>((m_buffer[2] << 8) | m_buffer[1]),
+								.raw = (length >= 3) ? m_buffer[3] : static_cast<uint8_t>(1),
+							}},
+							.type = ResourceData::Type::IRQ,
+						};
+						break;
+					case 0x05:
+						if (length < 2)
+							break;
+						result = ResourceData {
+							.as = { .dma = {
+								.channel_mask = m_buffer[1],
+								.raw = m_buffer[2],
+							}},
+							.type = ResourceData::Type::DMA,
+						};
+						break;
+					case 0x08:
+						if (length < 7)
+							break;
+						result = ResourceData {
+							.as = { .io_port = {
+								.range_min_base = static_cast<uint16_t>(((m_buffer[3] << 8) | m_buffer[2]) & ((m_buffer[1] & 1) ? 0xFFFF : 0x03FF)),
+								.range_max_base = static_cast<uint16_t>(((m_buffer[5] << 8) | m_buffer[4]) & ((m_buffer[1] & 1) ? 0xFFFF : 0x03FF)),
+								.base_alignment = m_buffer[6],
+								.range_length = m_buffer[7],
+							}},
+							.type = ResourceData::Type::IOPort,
+						};
+						break;
+					case 0x09:
+						if (length < 3)
+							break;
+						result = ResourceData {
+							.as = { .fixed_io_port = {
+								.range_base = static_cast<uint16_t>(((m_buffer[2] << 8) | m_buffer[1]) & 0x03FF),
+								.range_length = m_buffer[3],
+							}},
+							.type = ResourceData::Type::FixedIOPort,
+						};
+						break;
+					case 0x0A:
+						if (length < 5)
+							break;
+						result = ResourceData {
+							.as = { .fixed_dma = {
+								.request_line = static_cast<uint16_t>((m_buffer[2] << 8) | m_buffer[1]),
+								.channel      = static_cast<uint16_t>((m_buffer[4] << 8) | m_buffer[3]),
+								.transfer_width = m_buffer[5],
+							}},
+							.type = ResourceData::Type::FixedDMA,
+						};
+						break;
+					case 0x0F:
+						// End tag
+						return {};
+					case 0x06:
+					case 0x07:
+					case 0x0E:
+						dprintln("Skipping short resource 0x{2H}", (m_buffer[0] >> 3) & 0x0F);
+						break;
+				}
+
+				m_buffer = m_buffer.slice(1 + length);
+				if (result.has_value())
+					return result.release_value();
+			}
+		}
+
+	private:
+		BAN::ConstByteSpan m_buffer;
+	};
+
+}

kernel/include/kernel/API/SharedPage.h

@@ -0,0 +1,37 @@
+#pragma once
+
+#include <stdint.h>
+
+namespace Kernel::API
+{
+
+	enum SharedPageFeature : uint32_t
+	{
+		SPF_GETTIME = 1 << 0,
+	};
+
+	struct SharedPage
+	{
+		uint8_t __sequence[0x100];
+
+		uint32_t features;
+
+		struct
+		{
+			uint8_t shift;
+			uint64_t mult;
+			uint64_t realtime_seconds;
+		} gettime_shared;
+
+		struct
+		{
+			struct
+			{
+				uint32_t seq;
+				uint64_t last_ns;
+				uint64_t last_tsc;
+			} gettime_local;
+		} cpus[];
+	};
+
+}

kernel/include/kernel/API/Syscall.h

@@ -0,0 +1,34 @@
+#pragma once
+
+#include <BAN/MacroUtils.h>
+
+#if defined(__x86_64__)
+	#define _kas_instruction "syscall"
+	#define _kas_result rax
+	#define _kas_arguments rdi, rsi, rdx, r10, r8, r9
+	#define _kas_globbers rcx, rdx, rdi, rsi, r8, r9, r10, r11
+#elif defined(__i686__)
+	#define _kas_instruction "int $0xF0"
+	#define _kas_result eax
+	#define _kas_arguments eax, ebx, ecx, edx, esi, edi
+	#define _kas_globbers
+#endif
+
+#define _kas_argument_var(index, value) register long _kas_a##index asm(_ban_stringify(_ban_get(index, _kas_arguments))) = (long)value;
+#define _kas_dummy_var(index, value) register long _kas_d##index asm(#value);
+#define _kas_input(index, _) "r"(_kas_a##index)
+#define _kas_output(index, _) , "=r"(_kas_d##index)
+#define _kas_globber(_, value) #value
+
+#define _kas_syscall(...) ({ \
+		register long _kas_ret asm(_ban_stringify(_kas_result)); \
+		_ban_for_each(_kas_argument_var, __VA_ARGS__) \
+		_ban_for_each(_kas_dummy_var, _kas_globbers) \
+		asm volatile( \
+			_kas_instruction \
+			: "=r"(_kas_ret) _ban_for_each(_kas_output, _kas_globbers) \
+			: _ban_for_each_comma(_kas_input, __VA_ARGS__) \
+			: "cc", "memory"); \
+		(void)_kas_a0; /* require 1 argument */ \
+		_kas_ret; \
+	})

kernel/include/kernel/Audio/AC97/Controller.h

@@ -6,10 +6,10 @@
 namespace Kernel
 {
 
-	class AC97AudioController : public AudioController, public Interruptable
+	class AC97AudioController final : public AudioController, public Interruptable
 	{
 	public:
-		static BAN::ErrorOr<BAN::RefPtr<AC97AudioController>> create(PCI::Device& pci_device);
+		static BAN::ErrorOr<void> create(PCI::Device& pci_device);
 
 		void handle_irq() override;
 
@@ -19,16 +19,24 @@ namespace Kernel
 		uint32_t get_channels() const override { return 2; }
 		uint32_t get_sample_rate() const override { return 48000; }
 
+		uint32_t get_total_pins() const override { return 1; }
+		uint32_t get_current_pin() const override { return 0; }
+		BAN::ErrorOr<void> set_current_pin(uint32_t pin) override { if (pin != 0) return BAN::Error::from_errno(EINVAL); return {}; }
+
+		BAN::ErrorOr<void> set_volume_mdB(int32_t) override;
+
 	private:
 		AC97AudioController(PCI::Device& pci_device)
 			: m_pci_device(pci_device)
 		{ }
 
+		uint32_t get_volume_data() const;
+
 		BAN::ErrorOr<void> initialize();
 		BAN::ErrorOr<void> initialize_bld();
 		BAN::ErrorOr<void> initialize_interrupts();
 
-		void queue_samples_to_bld();
+		bool queue_samples_to_bld();
 
 	private:
 		static constexpr size_t m_bdl_entries = 32;

kernel/include/kernel/Audio/Controller.h

@@ -1,29 +1,39 @@
 #pragma once
 
 #include <kernel/Device/Device.h>
+#include <kernel/Memory/ByteRingBuffer.h>
 #include <kernel/PCI.h>
 
+#include <sys/ioctl.h>
+
 namespace Kernel
 {
 
 	class AudioController : public CharacterDevice
 	{
 	public:
-		static BAN::ErrorOr<BAN::RefPtr<AudioController>> create(PCI::Device& pci_device);
+		static BAN::ErrorOr<void> create(PCI::Device& pci_device);
 
 		dev_t rdev() const override { return m_rdev; }
 		BAN::StringView name() const override { return m_name; }
 
 	protected:
 		AudioController();
+		BAN::ErrorOr<void> initialize();
 
 		virtual void handle_new_data() = 0;
 
 		virtual uint32_t get_channels() const = 0;
 		virtual uint32_t get_sample_rate() const = 0;
 
+		virtual uint32_t get_total_pins() const = 0;
+		virtual uint32_t get_current_pin() const = 0;
+		virtual BAN::ErrorOr<void> set_current_pin(uint32_t) = 0;
+
+		virtual BAN::ErrorOr<void> set_volume_mdB(int32_t) = 0;
+
 		bool can_read_impl() const override { return false; }
-		bool can_write_impl() const override { SpinLockGuard _(m_spinlock); return m_sample_data_size < m_sample_data_capacity; }
+		bool can_write_impl() const override { SpinLockGuard _(m_spinlock); return !m_sample_data->full(); }
 		bool has_error_impl() const override { return false; }
 		bool has_hungup_impl() const override { return false; }
 
@@ -36,9 +46,9 @@ namespace Kernel
 		mutable SpinLock m_spinlock;
 
 		static constexpr size_t m_sample_data_capacity = 1 << 20;
-		uint8_t m_sample_data[m_sample_data_capacity];
+		BAN::UniqPtr<ByteRingBuffer> m_sample_data;
-		size_t m_sample_data_head { 0 };
+
-		size_t m_sample_data_size { 0 };
+		snd_volume_info m_volume_info {};
 
 	private:
 		const dev_t m_rdev;

kernel/include/kernel/Audio/HDAudio/AudioFunctionGroup.h

@@ -0,0 +1,80 @@
+#pragma once
+
+#include <kernel/Audio/Controller.h>
+#include <kernel/Audio/HDAudio/Controller.h>
+
+namespace Kernel
+{
+
+	class HDAudioController;
+
+	class HDAudioFunctionGroup final : public AudioController
+	{
+	public:
+		static BAN::ErrorOr<BAN::RefPtr<HDAudioFunctionGroup>> create(BAN::RefPtr<HDAudioController>, uint8_t cid, HDAudio::AFGNode&&);
+
+		void on_stream_interrupt(uint8_t stream_index);
+
+	protected:
+		// FIXME: allow setting these :D
+		uint32_t get_channels() const override { return 2; }
+		uint32_t get_sample_rate() const override { return 48000; }
+
+		uint32_t get_total_pins() const override;
+		uint32_t get_current_pin() const override;
+		BAN::ErrorOr<void> set_current_pin(uint32_t) override;
+
+		BAN::ErrorOr<void> set_volume_mdB(int32_t) override;
+
+		void handle_new_data() override;
+
+	private:
+		HDAudioFunctionGroup(BAN::RefPtr<HDAudioController> controller, uint8_t cid, HDAudio::AFGNode&& afg_node)
+			: m_controller(controller)
+			, m_afg_node(BAN::move(afg_node))
+			, m_cid(cid)
+		{ }
+		~HDAudioFunctionGroup();
+
+		BAN::ErrorOr<void> initialize();
+		BAN::ErrorOr<void> initialize_stream();
+		BAN::ErrorOr<void> initialize_output();
+
+		BAN::ErrorOr<void> enable_output_path(uint8_t index);
+		BAN::ErrorOr<void> disable_output_path(uint8_t index);
+
+		void reset_stream();
+
+		BAN::ErrorOr<void> recurse_output_paths(const HDAudio::AFGWidget& widget, BAN::Vector<const HDAudio::AFGWidget*>& path);
+
+		uint16_t get_format_data() const;
+
+		size_t bdl_offset() const;
+
+		void queue_bdl_data();
+
+	private:
+		// use 6 512 sample BDL entries
+		// each entry is ~10.7 ms at 48 kHz
+		// -> total buffered audio is 64 ms
+		static constexpr size_t m_bdl_entry_sample_frames = 512;
+		static constexpr size_t m_bdl_entry_count         = 6;
+
+		BAN::RefPtr<HDAudioController> m_controller;
+		const HDAudio::AFGNode m_afg_node;
+		const uint8_t m_cid;
+
+		BAN::Vector<BAN::Vector<const HDAudio::AFGWidget*>> m_output_paths;
+		BAN::Vector<const HDAudio::AFGWidget*> m_output_pins;
+		size_t m_output_path_index { SIZE_MAX };
+
+		uint8_t m_stream_id    { 0xFF };
+		uint8_t m_stream_index { 0xFF };
+		BAN::UniqPtr<DMARegion> m_bdl_region;
+
+		size_t m_bdl_head { 0 };
+		size_t m_bdl_tail { 0 };
+		bool m_stream_running { false };
+	};
+
+}

kernel/include/kernel/Audio/HDAudio/Controller.h

@@ -0,0 +1,77 @@
+#pragma once
+
+#include <kernel/Audio/Controller.h>
+#include <kernel/Audio/HDAudio/Definitions.h>
+#include <kernel/Memory/DMARegion.h>
+
+namespace Kernel
+{
+
+	class HDAudioController : public Interruptable, public BAN::RefCounted<HDAudioController>
+	{
+	public:
+		static BAN::ErrorOr<void> create(PCI::Device& pci_device);
+
+		BAN::ErrorOr<uint32_t> send_command(HDAudio::CORBEntry);
+
+		uint8_t get_stream_index(HDAudio::StreamType type, uint8_t index) const;
+
+		BAN::ErrorOr<uint8_t> allocate_stream_id();
+		void deallocate_stream_id(uint8_t id);
+
+		BAN::ErrorOr<uint8_t> allocate_stream(HDAudio::StreamType type, void* afg);
+		void deallocate_stream(uint8_t index);
+
+		PCI::BarRegion& bar0() { return *m_bar0; }
+
+		bool is_64bit() const { return m_is64bit; }
+
+		void handle_irq() override;
+
+	private:
+		HDAudioController(PCI::Device& pci_device)
+			: m_pci_device(pci_device)
+		{ }
+
+		BAN::ErrorOr<void> initialize();
+		BAN::ErrorOr<void> initialize_ring_buffers();
+
+		BAN::ErrorOr<void> reset_controller();
+
+		BAN::ErrorOr<HDAudio::Codec> initialize_codec(uint8_t codec);
+		BAN::ErrorOr<HDAudio::AFGNode> initialize_node(uint8_t codec, uint8_t node);
+		BAN::ErrorOr<HDAudio::AFGWidget> initialize_widget(uint8_t codec, uint8_t node);
+
+	private:
+		struct RingBuffer
+		{
+			vaddr_t vaddr;
+			uint32_t index;
+			uint32_t size;
+		};
+
+	private:
+		PCI::Device& m_pci_device;
+		BAN::UniqPtr<PCI::BarRegion> m_bar0;
+		bool m_is64bit { false };
+
+		bool m_use_immediate_command { false };
+
+		uint8_t m_output_streams { 0 };
+		uint8_t m_input_streams { 0 };
+		uint8_t m_bidir_streams { 0 };
+		void* m_allocated_streams[30] {};
+
+		// NOTE: stream ids are from 1 to 15
+		uint16_t m_allocated_stream_ids { 0 };
+
+		Mutex m_command_mutex;
+		SpinLock m_rb_lock;
+		ThreadBlocker m_rb_blocker;
+
+		RingBuffer m_corb;
+		RingBuffer m_rirb;
+		BAN::UniqPtr<DMARegion> m_ring_buffer_region;
+	};
+
+}

kernel/include/kernel/Audio/HDAudio/Definitions.h

@@ -0,0 +1,90 @@
+#pragma once
+
+#include <BAN/Vector.h>
+
+namespace Kernel::HDAudio
+{
+
+	struct CORBEntry
+	{
+		union {
+			struct {
+				uint32_t data          : 8;
+				uint32_t command       : 12;
+				uint32_t node_index    : 8;
+				uint32_t codec_address : 4;
+			};
+			uint32_t raw;
+		};
+	};
+	static_assert(sizeof(CORBEntry) == sizeof(uint32_t));
+
+	struct BDLEntry
+	{
+		paddr_t address;
+		uint32_t length;
+		uint32_t ioc;
+	};
+	static_assert(sizeof(BDLEntry) == 16);
+
+	struct AFGWidget
+	{
+		enum class Type
+		{
+			OutputConverter,
+			InputConverter,
+			Mixer,
+			Selector,
+			PinComplex,
+			Power,
+			VolumeKnob,
+			BeepGenerator,
+		};
+
+		Type type;
+		uint8_t id;
+
+		union
+		{
+			struct
+			{
+				bool input;
+				bool output;
+				bool display; // HDMI or DP
+				uint32_t config;
+			} pin_complex;
+		};
+
+		struct Amplifier
+		{
+			uint8_t offset;
+			uint8_t num_steps;
+			uint8_t step_size;
+			bool mute;
+		};
+
+		BAN::Optional<Amplifier> output_amplifier;
+
+		BAN::Vector<uint16_t> connections;
+	};
+
+	struct AFGNode
+	{
+		uint8_t id;
+		BAN::Vector<AFGWidget> widgets;
+	};
+
+	struct Codec
+	{
+		uint8_t id;
+		BAN::Vector<AFGNode> nodes;
+	};
+
+	enum class StreamType
+	{
+		Input,
+		Output,
+		Bidirectional,
+	};
+
+}

kernel/include/kernel/Audio/HDAudio/Registers.h

@@ -0,0 +1,50 @@
+#pragma once
+
+#include <stdint.h>
+
+namespace Kernel::HDAudio
+{
+
+	enum Regs : uint8_t
+	{
+		GCAP = 0x00,
+		VMIN = 0x02,
+		VMAJ = 0x03,
+		GCTL = 0x08,
+		STATESTS = 0x0E,
+
+		INTCTL = 0x20,
+		INTSTS = 0x24,
+
+		CORBLBASE = 0x40,
+		CORBUBASE = 0x44,
+		CORBWP    = 0x48,
+		CORBRP    = 0x4A,
+		CORBCTL   = 0x4C,
+		CORBSTS   = 0x4D,
+		CORBSIZE  = 0x4E,
+
+		RIRBLBASE = 0x50,
+		RIRBUBASE = 0x54,
+		RIRBWP    = 0x58,
+		RINTCNT   = 0x5A,
+		RIRBCTL   = 0x5C,
+		RIRBSTS   = 0x5D,
+		RIRBSIZE  = 0x5E,
+
+		ICOI = 0x60,
+		ICII = 0x64,
+		ICIS = 0x68,
+
+		SDCTL   = 0x00,
+		SDSTS   = 0x03,
+		SDLPIB  = 0x04,
+		SDCBL   = 0x08,
+		SDLVI   = 0x0C,
+		SDFIFOD = 0x10,
+		SDFMT   = 0x12,
+		SDBDPL  = 0x18,
+		SDBDPU  = 0x1C,
+	};
+
+}

kernel/include/kernel/BootInfo.h

@@ -44,7 +44,7 @@ namespace Kernel
 	struct BootModule
 	{
 		paddr_t start;
-		size_t size;
+		uint64_t size;
 	};
 
 	struct BootInfo

kernel/include/kernel/CPUID.h

@@ -81,5 +81,6 @@ namespace CPUID
 	bool has_pge();
 	bool has_pat();
 	bool has_1gib_pages();
+	bool has_invariant_tsc();
 
 }

kernel/include/kernel/Credentials.h

@@ -35,10 +35,8 @@ namespace Kernel
 
 		bool has_egid(gid_t) const;
 
-		BAN::ErrorOr<void> initialize_supplementary_groups();
+		BAN::Span<const gid_t> groups() const { return m_supplementary.span(); }
-
+		BAN::ErrorOr<void> set_groups(BAN::Span<const gid_t> groups);
-	private:
-		BAN::ErrorOr<BAN::String> find_username() const;
 
 	private:
 		uid_t m_ruid, m_euid, m_suid;

kernel/include/kernel/Debug.h

@@ -70,10 +70,15 @@
 #define DEBUG_USB_MOUSE 0
 #define DEBUG_USB_MASS_STORAGE 0
 
+#define DEBUG_HDAUDIO 0
+
 
 namespace Debug
 {
 	void dump_stack_trace();
+	void dump_stack_trace(uintptr_t ip, uintptr_t bp);
+	void dump_qr_code();
+
 	void putchar(char);
 	void print_prefix(const char*, int);
 

kernel/include/kernel/Device/Device.h

@@ -16,7 +16,11 @@ namespace Kernel
 		virtual bool is_partition() const { return false; }
 		virtual bool is_storage_device() const { return false; }
 
-		virtual BAN::ErrorOr<BAN::UniqPtr<MemoryRegion>> mmap_region(PageTable&, off_t offset, size_t len, AddressRange, MemoryRegion::Type, PageTable::flags_t) { (void)offset; (void)len; return BAN::Error::from_errno(EINVAL); }
+		virtual BAN::ErrorOr<BAN::UniqPtr<MemoryRegion>> mmap_region(PageTable&, off_t offset, size_t len, AddressRange, MemoryRegion::Type, PageTable::flags_t, int status_flags)
+		{
+			(void)offset; (void)len; (void)status_flags;
+			return BAN::Error::from_errno(ENOTSUP);
+		}
 
 		virtual dev_t rdev() const override = 0;
 

kernel/include/kernel/Device/DeviceNumbers.h

@@ -16,6 +16,7 @@ namespace Kernel
 		Debug,
 		Keyboard,
 		Mouse,
+		Joystick,
 		SCSI,
 		NVMeController,
 		NVMeNamespace,

kernel/include/kernel/Device/FramebufferDevice.h

@@ -10,6 +10,7 @@ namespace Kernel
 	{
 	public:
 		static BAN::ErrorOr<BAN::RefPtr<FramebufferDevice>> create_from_boot_framebuffer();
+		static BAN::RefPtr<FramebufferDevice> boot_framebuffer();
 		~FramebufferDevice();
 
 		uint32_t width() const { return m_width; }
@@ -27,7 +28,7 @@ namespace Kernel
 		void sync_pixels_linear(uint32_t first_pixel, uint32_t pixel_count);
 		void sync_pixels_rectangle(uint32_t top_right_x, uint32_t top_right_y, uint32_t width, uint32_t height);
 
-		virtual BAN::ErrorOr<BAN::UniqPtr<MemoryRegion>> mmap_region(PageTable&, off_t offset, size_t len, AddressRange, MemoryRegion::Type, PageTable::flags_t) override;
+		virtual BAN::ErrorOr<BAN::UniqPtr<MemoryRegion>> mmap_region(PageTable&, off_t offset, size_t len, AddressRange, MemoryRegion::Type, PageTable::flags_t, int status_flags) override;
 
 		virtual dev_t rdev() const override { return m_rdev; }
 		virtual BAN::StringView name() const override { return m_name.sv(); }
@@ -36,6 +37,8 @@ namespace Kernel
 		virtual BAN::ErrorOr<size_t> read_impl(off_t, BAN::ByteSpan) override;
 		virtual BAN::ErrorOr<size_t> write_impl(off_t, BAN::ConstByteSpan) override;
 
+		BAN::ErrorOr<long> ioctl_impl(int cmd, void* arg) override;
+
 		virtual bool can_read_impl() const override { return true; }
 		virtual bool can_write_impl() const override { return true; }
 		virtual bool has_error_impl() const override { return false; }

kernel/include/kernel/ELF.h

@@ -20,6 +20,6 @@ namespace Kernel::ELF
 		BAN::Vector<BAN::UniqPtr<MemoryRegion>> regions;
 	};
 
-	BAN::ErrorOr<LoadResult> load_from_inode(BAN::RefPtr<Inode>, const Credentials&, PageTable&);
+	BAN::ErrorOr<LoadResult> load_from_inode(BAN::RefPtr<Inode> root, BAN::RefPtr<Inode> inode, const Credentials&, PageTable&);
 
 }

kernel/include/kernel/FS/DevFS/FileSystem.h

@@ -22,6 +22,7 @@ namespace Kernel
 
 		void add_inode(BAN::StringView path, BAN::RefPtr<TmpInode>);
 
+		void initiate_disk_cache_drop();
 		void initiate_sync(bool should_block);
 
 	private:
@@ -37,6 +38,10 @@ namespace Kernel
 		ThreadBlocker m_sync_done;
 		ThreadBlocker m_sync_thread_blocker;
 		volatile bool m_should_sync { false };
+
+		SpinLock m_disk_cache_lock;
+		ThreadBlocker m_disk_cache_thread_blocker;
+		BAN::Atomic<bool> m_should_drop_disk_cache { false };
 	};
 
 }

kernel/include/kernel/FS/EventFD.h

@@ -0,0 +1,52 @@
+#pragma once
+
+#include <kernel/FS/Inode.h>
+
+namespace Kernel
+{
+
+	class EventFD final : public Inode
+	{
+	public:
+		static BAN::ErrorOr<BAN::RefPtr<Inode>> create(uint64_t initval, bool semaphore);
+
+		ino_t ino() const override { return 0; }
+		Mode mode() const override { return { Mode::IFCHR | Mode::IRUSR | Mode::IWUSR }; }
+		nlink_t nlink() const override { return ref_count(); }
+		uid_t uid() const override { return 0; }
+		gid_t gid() const override { return 0; }
+		off_t size() const override { return 0; }
+		timespec atime() const override { return {}; }
+		timespec mtime() const override { return {}; }
+		timespec ctime() const override { return {}; }
+		blksize_t blksize() const override { return 8; }
+		blkcnt_t blocks() const override { return 0; }
+		dev_t dev() const override { return 0; }
+		dev_t rdev() const override { return 0; }
+
+		const FileSystem* filesystem() const override { return nullptr; }
+
+	protected:
+		BAN::ErrorOr<size_t> read_impl(off_t, BAN::ByteSpan) override;
+		BAN::ErrorOr<size_t> write_impl(off_t, BAN::ConstByteSpan) override;
+		BAN::ErrorOr<void> fsync_impl() final override { return {}; }
+
+		bool can_read_impl() const override { return m_value > 0; }
+		bool can_write_impl() const override { return m_value < UINT64_MAX - 1; }
+		bool has_error_impl() const override { return false; }
+		bool has_hungup_impl() const override { return false; }
+
+	private:
+		EventFD(uint64_t initval, bool is_semaphore)
+			: m_is_semaphore(is_semaphore)
+			, m_value(initval)
+		{ }
+
+	private:
+		const bool m_is_semaphore;
+		uint64_t m_value;
+
+		ThreadBlocker m_thread_blocker;
+	};
+
+}

kernel/include/kernel/FS/Ext2/FileSystem.h

@@ -26,18 +26,32 @@ namespace Kernel
 		class BlockBufferWrapper
 		{
 			BAN_NON_COPYABLE(BlockBufferWrapper);
-			BAN_NON_MOVABLE(BlockBufferWrapper);
 
 		public:
-			BlockBufferWrapper(BAN::Span<uint8_t> buffer, bool& used)
+			BlockBufferWrapper(BAN::Span<uint8_t> buffer, void (*callback)(void*, const uint8_t*), void* argument)
 				: m_buffer(buffer)
-				, m_used(used)
+				, m_callback(callback)
-			{
+				, m_argument(argument)
-				ASSERT(m_used);
+			{ }
-			}
+			BlockBufferWrapper(BlockBufferWrapper&& other) { *this = BAN::move(other); }
 			~BlockBufferWrapper()
 			{
-				m_used = false;
+				if (m_callback == nullptr)
+					return;
+				m_callback(m_argument, m_buffer.data());
+			}
+
+			BlockBufferWrapper& operator=(BlockBufferWrapper&& other)
+			{
+				this->m_buffer = other.m_buffer;
+				this->m_callback = other.m_callback;
+				this->m_argument = other.m_argument;
+
+				other.m_buffer = {};
+				other.m_callback = nullptr;
+				other.m_argument = nullptr;
+
+				return *this;
 			}
 
 			size_t size() const { return m_buffer.size(); }
@@ -53,7 +67,8 @@ namespace Kernel
 
 		private:
 			BAN::Span<uint8_t> m_buffer;
-			bool& m_used;
+			void (*m_callback)(void*, const uint8_t*);
+			void* m_argument;
 		};
 
 	public:
@@ -79,7 +94,7 @@ namespace Kernel
 		BAN::ErrorOr<void> sync_superblock();
 		BAN::ErrorOr<void> sync_block(uint32_t block);
 
-		BlockBufferWrapper get_block_buffer();
+		BAN::ErrorOr<BlockBufferWrapper> get_block_buffer();
 
 		BAN::ErrorOr<uint32_t> reserve_free_block(uint32_t primary_bgd);
 		BAN::ErrorOr<void> release_block(uint32_t block);
@@ -102,10 +117,13 @@ namespace Kernel
 		{
 		public:
 			BlockBufferManager() = default;
-			BlockBufferWrapper get_buffer();
+			BAN::ErrorOr<BlockBufferWrapper> get_buffer();
 
 			BAN::ErrorOr<void> initialize(size_t block_size);
 
+		private:
+			void destroy_callback(const uint8_t* buffer_ptr);
+
 		private:
 			struct BlockBuffer
 			{
@@ -113,8 +131,20 @@ namespace Kernel
 				bool used { false };
 			};
 
+			struct ThreadInfo
+			{
+				pid_t tid { 0 };
+				size_t buffers { 0 };
+			};
+
 		private:
-			BAN::Array<BlockBuffer, 10> m_buffers;
+			static constexpr size_t max_threads = 8;
+			static constexpr size_t max_buffers_per_thread = 6;
+
+			Mutex m_buffer_mutex;
+			ThreadBlocker m_buffer_blocker;
+			BAN::Array<BlockBuffer, max_threads * max_buffers_per_thread> m_buffers;
+			BAN::Array<ThreadInfo, max_threads> m_thread_infos;
 		};
 
 	private:

kernel/include/kernel/FS/Ext2/Inode.h

@@ -37,6 +37,7 @@ namespace Kernel
 		virtual BAN::ErrorOr<void> create_file_impl(BAN::StringView, mode_t, uid_t, gid_t) override;
 		virtual BAN::ErrorOr<void> create_directory_impl(BAN::StringView, mode_t, uid_t, gid_t) override;
 		virtual BAN::ErrorOr<void> link_inode_impl(BAN::StringView, BAN::RefPtr<Inode>) override;
+		virtual BAN::ErrorOr<void> rename_inode_impl(BAN::RefPtr<Inode>, BAN::StringView, BAN::StringView) override;
 		virtual BAN::ErrorOr<void> unlink_impl(BAN::StringView) override;
 
 		virtual BAN::ErrorOr<BAN::String> link_target_impl() override;
@@ -60,18 +61,18 @@ namespace Kernel
 		// NOTE: the inode might have more blocks than what this suggests if it has been shrinked
 		uint32_t max_used_data_block_count() const { return size() / blksize(); }
 
-		BAN::ErrorOr<BAN::Optional<uint32_t>> block_from_indirect_block(uint32_t block, uint32_t index, uint32_t depth);
+		BAN::ErrorOr<BAN::Optional<uint32_t>> block_from_indirect_block(uint32_t& block, uint32_t index, uint32_t depth, bool allocate);
-		BAN::ErrorOr<BAN::Optional<uint32_t>> fs_block_of_data_block_index(uint32_t data_block_index);
+		BAN::ErrorOr<BAN::Optional<uint32_t>> fs_block_of_data_block_index(uint32_t data_block_index, bool allocate);
 
 		BAN::ErrorOr<void> link_inode_to_directory(Ext2Inode&, BAN::StringView name);
+		BAN::ErrorOr<void> remove_inode_from_directory(BAN::StringView name, bool cleanup_directory);
 		BAN::ErrorOr<bool> is_directory_empty();
 
 		BAN::ErrorOr<void> cleanup_indirect_block(uint32_t block, uint32_t depth);
 		BAN::ErrorOr<void> cleanup_default_links();
+		BAN::ErrorOr<void> cleanup_data_blocks();
 		BAN::ErrorOr<void> cleanup_from_fs();
 
-		BAN::ErrorOr<uint32_t> allocate_new_block_to_indirect_block(uint32_t& block, uint32_t index, uint32_t depth);
-		BAN::ErrorOr<uint32_t> allocate_new_block(uint32_t data_block_index);
 		BAN::ErrorOr<void> sync();
 
 		uint32_t block_group() const;
@@ -84,6 +85,26 @@ namespace Kernel
 		{}
 		static BAN::ErrorOr<BAN::RefPtr<Ext2Inode>> create(Ext2FS&, uint32_t);
 
+	private:
+		struct ScopedSync
+		{
+			ScopedSync(Ext2Inode& inode)
+				: inode(inode)
+				, inode_info(inode.m_inode)
+			{ }
+
+			~ScopedSync()
+			{
+				if (memcmp(&inode.m_inode, &inode_info, sizeof(Ext2::Inode)) == 0)
+					return;
+				if (auto ret = inode.sync(); ret.is_error())
+					dwarnln("failed to sync inode: {}", ret.error());
+			}
+
+			Ext2Inode& inode;
+			Ext2::Inode inode_info;
+		};
+
 	private:
 		Ext2FS& m_fs;
 		Ext2::Inode m_inode;

kernel/include/kernel/FS/Inode.h

@@ -22,7 +22,7 @@ namespace Kernel
 
 	class FileBackedRegion;
 	class FileSystem;
-	class SharedFileData;
+	struct SharedFileData;
 
 	class Inode : public BAN::RefCounted<Inode>
 	{
@@ -97,6 +97,7 @@ namespace Kernel
 		BAN::ErrorOr<void> create_file(BAN::StringView, mode_t, uid_t, gid_t);
 		BAN::ErrorOr<void> create_directory(BAN::StringView, mode_t, uid_t, gid_t);
 		BAN::ErrorOr<void> link_inode(BAN::StringView, BAN::RefPtr<Inode>);
+		BAN::ErrorOr<void> rename_inode(BAN::RefPtr<Inode>, BAN::StringView, BAN::StringView);
 		BAN::ErrorOr<void> unlink(BAN::StringView);
 
 		// Link API
@@ -108,10 +109,12 @@ namespace Kernel
 		BAN::ErrorOr<void> bind(const sockaddr* address, socklen_t address_len);
 		BAN::ErrorOr<void> connect(const sockaddr* address, socklen_t address_len);
 		BAN::ErrorOr<void> listen(int backlog);
-		BAN::ErrorOr<size_t> sendto(BAN::ConstByteSpan message, const sockaddr* address, socklen_t address_len);
+		BAN::ErrorOr<size_t> sendmsg(const msghdr& message, int flags);
-		BAN::ErrorOr<size_t> recvfrom(BAN::ByteSpan buffer, sockaddr* address, socklen_t* address_len);
+		BAN::ErrorOr<size_t> recvmsg(msghdr& message, int flags);
 		BAN::ErrorOr<void> getsockname(sockaddr* address, socklen_t* address_len);
 		BAN::ErrorOr<void> getpeername(sockaddr* address, socklen_t* address_len);
+		BAN::ErrorOr<void> getsockopt(int level, int option, void* value, socklen_t* value_len);
+		BAN::ErrorOr<void> setsockopt(int level, int option, const void* value, socklen_t value_len);
 
 		// General API
 		BAN::ErrorOr<size_t> read(off_t, BAN::ByteSpan buffer);
@@ -139,12 +142,13 @@ namespace Kernel
 
 	protected:
 		// Directory API
-		virtual BAN::ErrorOr<BAN::RefPtr<Inode>> find_inode_impl(BAN::StringView)				{ return BAN::Error::from_errno(ENOTSUP); }
+		virtual BAN::ErrorOr<BAN::RefPtr<Inode>> find_inode_impl(BAN::StringView)							{ return BAN::Error::from_errno(ENOTSUP); }
-		virtual BAN::ErrorOr<size_t> list_next_inodes_impl(off_t, struct dirent*, size_t)		{ return BAN::Error::from_errno(ENOTSUP); }
+		virtual BAN::ErrorOr<size_t> list_next_inodes_impl(off_t, struct dirent*, size_t)					{ return BAN::Error::from_errno(ENOTSUP); }
-		virtual BAN::ErrorOr<void> create_file_impl(BAN::StringView, mode_t, uid_t, gid_t)		{ return BAN::Error::from_errno(ENOTSUP); }
+		virtual BAN::ErrorOr<void> create_file_impl(BAN::StringView, mode_t, uid_t, gid_t)					{ return BAN::Error::from_errno(ENOTSUP); }
-		virtual BAN::ErrorOr<void> create_directory_impl(BAN::StringView, mode_t, uid_t, gid_t)	{ return BAN::Error::from_errno(ENOTSUP); }
+		virtual BAN::ErrorOr<void> create_directory_impl(BAN::StringView, mode_t, uid_t, gid_t)				{ return BAN::Error::from_errno(ENOTSUP); }
-		virtual BAN::ErrorOr<void> link_inode_impl(BAN::StringView, BAN::RefPtr<Inode>)			{ return BAN::Error::from_errno(ENOTSUP); }
+		virtual BAN::ErrorOr<void> link_inode_impl(BAN::StringView, BAN::RefPtr<Inode>)						{ return BAN::Error::from_errno(ENOTSUP); }
-		virtual BAN::ErrorOr<void> unlink_impl(BAN::StringView)									{ return BAN::Error::from_errno(ENOTSUP); }
+		virtual BAN::ErrorOr<void> rename_inode_impl(BAN::RefPtr<Inode>, BAN::StringView, BAN::StringView)	{ return BAN::Error::from_errno(ENOTSUP); }
+		virtual BAN::ErrorOr<void> unlink_impl(BAN::StringView)												{ return BAN::Error::from_errno(ENOTSUP); }
 
 		// Link API
 		virtual BAN::ErrorOr<BAN::String> link_target_impl()				{ return BAN::Error::from_errno(ENOTSUP); }
@@ -155,10 +159,12 @@ namespace Kernel
 		virtual BAN::ErrorOr<void> connect_impl(const sockaddr*, socklen_t)							{ return BAN::Error::from_errno(ENOTSUP); }
 		virtual BAN::ErrorOr<void> listen_impl(int)													{ return BAN::Error::from_errno(ENOTSUP); }
 		virtual BAN::ErrorOr<void> bind_impl(const sockaddr*, socklen_t)							{ return BAN::Error::from_errno(ENOTSUP); }
-		virtual BAN::ErrorOr<size_t> sendto_impl(BAN::ConstByteSpan, const sockaddr*, socklen_t)	{ return BAN::Error::from_errno(ENOTSUP); }
+		virtual BAN::ErrorOr<size_t> recvmsg_impl(msghdr&, int)										{ return BAN::Error::from_errno(ENOTSUP); }
-		virtual BAN::ErrorOr<size_t> recvfrom_impl(BAN::ByteSpan, sockaddr*, socklen_t*)			{ return BAN::Error::from_errno(ENOTSUP); }
+		virtual BAN::ErrorOr<size_t> sendmsg_impl(const msghdr&, int)								{ return BAN::Error::from_errno(ENOTSUP); }
 		virtual BAN::ErrorOr<void> getsockname_impl(sockaddr*, socklen_t*)							{ return BAN::Error::from_errno(ENOTSUP); }
 		virtual BAN::ErrorOr<void> getpeername_impl(sockaddr*, socklen_t*)							{ return BAN::Error::from_errno(ENOTSUP); }
+		virtual BAN::ErrorOr<void> getsockopt_impl(int, int, void*, socklen_t*)						{ return BAN::Error::from_errno(ENOTSUP); }
+		virtual BAN::ErrorOr<void> setsockopt_impl(int, int, const void*, socklen_t)				{ return BAN::Error::from_errno(ENOTSUP); }
 
 		// General API
 		virtual BAN::ErrorOr<size_t> read_impl(off_t, BAN::ByteSpan)		{ return BAN::Error::from_errno(ENOTSUP); }
@@ -187,7 +193,7 @@ namespace Kernel
 		friend class Epoll;
 		friend class FileBackedRegion;
 		friend class OpenFileDescriptorSet;
-		friend class SharedFileData;
+		friend struct SharedFileData;
 		friend class TTY;
 	};
 

kernel/include/kernel/FS/Pipe.h

@@ -2,6 +2,7 @@
 
 #include <BAN/Array.h>
 #include <kernel/FS/Inode.h>
+#include <kernel/Memory/ByteRingBuffer.h>
 #include <kernel/ThreadBlocker.h>
 
 namespace Kernel
@@ -38,7 +39,7 @@ namespace Kernel
 		virtual BAN::ErrorOr<size_t> write_impl(off_t, BAN::ConstByteSpan) override;
 		virtual BAN::ErrorOr<void> fsync_impl() final override { return {}; }
 
-		virtual bool can_read_impl() const override { return m_buffer_size > 0; }
+		virtual bool can_read_impl() const override { return !m_buffer->empty(); }
 		virtual bool can_write_impl() const override { return true; }
 		virtual bool has_error_impl() const override { return m_reading_count == 0; }
 		virtual bool has_hungup_impl() const override { return m_writing_count == 0; }
@@ -54,9 +55,7 @@ namespace Kernel
 		timespec m_ctime {};
 		ThreadBlocker m_thread_blocker;
 
-		BAN::Array<uint8_t, PAGE_SIZE> m_buffer;
+		BAN::UniqPtr<ByteRingBuffer> m_buffer;
-		BAN::Atomic<size_t> m_buffer_size { 0 };
-		size_t m_buffer_tail { 0 };
 
 		BAN::Atomic<uint32_t> m_writing_count { 1 };
 		BAN::Atomic<uint32_t> m_reading_count { 1 };

kernel/include/kernel/FS/ProcFS/Inode.h

@@ -57,6 +57,31 @@ namespace Kernel
 		size_t (Process::*m_callback)(off_t, BAN::ByteSpan) const;
 	};
 
+	class ProcSymlinkProcessInode final : public TmpInode
+	{
+	public:
+		static BAN::ErrorOr<BAN::RefPtr<ProcSymlinkProcessInode>> create_new(Process& process, BAN::ErrorOr<BAN::String> (Process::*callback)() const, TmpFileSystem&, mode_t);
+		~ProcSymlinkProcessInode() = default;
+
+		virtual uid_t uid() const override { return m_process.credentials().ruid(); }
+		virtual gid_t gid() const override { return m_process.credentials().rgid(); }
+
+	protected:
+		virtual BAN::ErrorOr<BAN::String> link_target_impl() override;
+
+		virtual bool can_read_impl() const override { return false; }
+		virtual bool can_write_impl() const override { return false; }
+		virtual bool has_error_impl() const override { return false; }
+		virtual bool has_hungup_impl() const override { return false; }
+
+	private:
+		ProcSymlinkProcessInode(Process& process, BAN::ErrorOr<BAN::String> (Process::*)() const, TmpFileSystem&, const TmpInodeInfo&);
+
+	private:
+		Process& m_process;
+		BAN::ErrorOr<BAN::String> (Process::*m_callback)() const;
+	};
+
 	class ProcROInode final : public TmpInode
 	{
 	public:
@@ -82,4 +107,57 @@ namespace Kernel
 		size_t (*m_callback)(off_t, BAN::ByteSpan);
 	};
 
+	class ProcSymlinkInode final : public TmpInode
+	{
+	public:
+		static BAN::ErrorOr<BAN::RefPtr<ProcSymlinkInode>> create_new(BAN::ErrorOr<BAN::String> (*)(void*), void (*)(void*), void* data, TmpFileSystem&, mode_t, uid_t, gid_t);
+		~ProcSymlinkInode();
+
+	protected:
+		virtual BAN::ErrorOr<BAN::String> link_target_impl() override;
+
+		virtual bool can_read_impl() const override { return false; }
+		virtual bool can_write_impl() const override { return false; }
+		virtual bool has_error_impl() const override { return false; }
+		virtual bool has_hungup_impl() const override { return false; }
+
+	private:
+		ProcSymlinkInode(BAN::ErrorOr<BAN::String> (*callback)(void*), void (*destructor)(void*), void* data, TmpFileSystem&, const TmpInodeInfo&);
+
+	private:
+		BAN::ErrorOr<BAN::String> (*m_callback)(void*);
+		void (*m_destructor)(void*);
+		void* m_data;
+	};
+
+	class ProcFDDirectoryInode final : public TmpInode
+	{
+	public:
+		static BAN::ErrorOr<BAN::RefPtr<ProcFDDirectoryInode>> create_new(Process&, TmpFileSystem&, mode_t);
+		~ProcFDDirectoryInode() = default;
+
+		virtual uid_t uid() const override { return m_process.credentials().ruid(); }
+		virtual gid_t gid() const override { return m_process.credentials().rgid(); }
+
+	protected:
+		virtual BAN::ErrorOr<BAN::RefPtr<Inode>> find_inode_impl(BAN::StringView) override;
+		virtual BAN::ErrorOr<size_t> list_next_inodes_impl(off_t, struct dirent*, size_t) override;
+		virtual BAN::ErrorOr<void> create_file_impl(BAN::StringView, mode_t, uid_t, gid_t) override                 { return BAN::Error::from_errno(EPERM); }
+		virtual BAN::ErrorOr<void> create_directory_impl(BAN::StringView, mode_t, uid_t, gid_t) override            { return BAN::Error::from_errno(EPERM); }
+		virtual BAN::ErrorOr<void> link_inode_impl(BAN::StringView, BAN::RefPtr<Inode>) override                    { return BAN::Error::from_errno(EPERM); }
+		virtual BAN::ErrorOr<void> rename_inode_impl(BAN::RefPtr<Inode>, BAN::StringView, BAN::StringView) override { return BAN::Error::from_errno(EPERM); }
+		virtual BAN::ErrorOr<void> unlink_impl(BAN::StringView) override                                            { return BAN::Error::from_errno(EPERM); }
+
+		virtual bool can_read_impl() const override { return false; }
+		virtual bool can_write_impl() const override { return false; }
+		virtual bool has_error_impl() const override { return false; }
+		virtual bool has_hungup_impl() const override { return false; }
+
+	private:
+		ProcFDDirectoryInode(Process&, TmpFileSystem&, const TmpInodeInfo&);
+
+	private:
+		Process& m_process;
+	};
+
 }

kernel/include/kernel/FS/Socket.h

@@ -51,8 +51,6 @@ namespace Kernel
 			: m_info(info)
 		{}
 
-		BAN::ErrorOr<size_t> read_impl(off_t, BAN::ByteSpan buffer) override { return recvfrom_impl(buffer, nullptr, nullptr); }
-		BAN::ErrorOr<size_t> write_impl(off_t, BAN::ConstByteSpan buffer) override { return sendto_impl(buffer, nullptr, 0); }
 		BAN::ErrorOr<void> fsync_impl() final override { return {}; }
 
 	private:

kernel/include/kernel/FS/TmpFS/Inode.h

@@ -156,6 +156,8 @@ namespace Kernel
 		virtual BAN::ErrorOr<size_t> list_next_inodes_impl(off_t, struct dirent*, size_t) override final;
 		virtual BAN::ErrorOr<void> create_file_impl(BAN::StringView, mode_t, uid_t, gid_t) override final;
 		virtual BAN::ErrorOr<void> create_directory_impl(BAN::StringView, mode_t, uid_t, gid_t) override final;
+		virtual BAN::ErrorOr<void> link_inode_impl(BAN::StringView, BAN::RefPtr<Inode>) override final;
+		virtual BAN::ErrorOr<void> rename_inode_impl(BAN::RefPtr<Inode>, BAN::StringView, BAN::StringView) override final;
 		virtual BAN::ErrorOr<void> unlink_impl(BAN::StringView) override;
 
 		virtual bool can_read_impl() const override { return false; }
@@ -167,6 +169,8 @@ namespace Kernel
 		template<TmpFuncs::for_each_valid_entry_callback F>
 		void for_each_valid_entry(F callback);
 
+		BAN::ErrorOr<void> unlink_inode(BAN::StringView, bool cleanup);
+
 		friend class TmpInode;
 	};
 

kernel/include/kernel/FS/USTARModule.h

@@ -1,12 +1,11 @@
 #pragma once
 
 #include <kernel/BootInfo.h>
-#include <kernel/FS/FileSystem.h>
+#include <kernel/FS/Inode.h>
 
 namespace Kernel
 {
 
-	bool is_ustar_boot_module(const BootModule&);
+	BAN::ErrorOr<bool> unpack_boot_module_into_directory(BAN::RefPtr<Inode>, const BootModule&);
-	BAN::ErrorOr<void> unpack_boot_module_into_filesystem(BAN::RefPtr<FileSystem>, const BootModule&);
 
 }

kernel/include/kernel/FS/VirtualFileSystem.h

@@ -71,13 +71,13 @@ namespace Kernel
 
 		File root_file()
 		{
-			return File(root_inode(), "/"_sv);
+			return File { root_inode(), "/"_sv };
 		}
 
-		BAN::ErrorOr<File> file_from_relative_path(const File& parent, const Credentials&, BAN::StringView, int);
+		BAN::ErrorOr<File> file_from_relative_path(BAN::RefPtr<Inode> root_inode, const File& parent, const Credentials&, BAN::StringView, int);
-		BAN::ErrorOr<File> file_from_absolute_path(const Credentials& credentials, BAN::StringView path, int flags)
+		BAN::ErrorOr<File> file_from_absolute_path(BAN::RefPtr<Inode> root_inode, const Credentials& credentials, BAN::StringView path, int flags)
 		{
-			return file_from_relative_path(root_file(), credentials, path, flags);
+			return file_from_relative_path(root_inode, File { root_inode, "/"_sv }, credentials, path, flags);
 		}
 
 	private:

kernel/include/kernel/GDT.h

@@ -129,7 +129,8 @@ namespace Kernel
 		}
 
 #if ARCH(i686)
-		void set_tls(uintptr_t addr);
+		void set_fsbase(uintptr_t addr);
+		void set_gsbase(uintptr_t addr);
 #endif
 
 	private:
@@ -150,11 +151,11 @@ namespace Kernel
 
 	private:
 #if ARCH(x86_64)
-		BAN::Array<SegmentDescriptor, 7> m_gdt; // null, kernel code, kernel data, user code, user data, tss low, tss high
+		BAN::Array<SegmentDescriptor, 8> m_gdt; // null, kernel code, kernel data, user code (32 bit), user data, user code (64 bit), tss low, tss high
-		static constexpr uint16_t m_tss_offset = 0x28;
+		static constexpr uint16_t m_tss_offset = 0x30;
 #elif ARCH(i686)
-		BAN::Array<SegmentDescriptor, 8> m_gdt; // null, kernel code, kernel data, user code, user data, processor data, tls, tss
+		BAN::Array<SegmentDescriptor, 9> m_gdt; // null, kernel code, kernel data, user code, user data, processor data, fsbase, gsbase, tss
-		static constexpr uint16_t m_tss_offset = 0x38;
+		static constexpr uint16_t m_tss_offset = 0x40;
 #endif
 		TaskStateSegment m_tss;
 		const GDTR m_gdtr {

kernel/include/kernel/IDT.h

@@ -18,10 +18,12 @@ namespace Kernel
 
 	constexpr uint8_t IRQ_VECTOR_BASE   = 0x20;
 	constexpr uint8_t IRQ_MSI_BASE      = 0x80;
-	constexpr uint8_t IRQ_SYSCALL       = 0xF0;
+	constexpr uint8_t IRQ_MSI_END       = 0xF0;
-	constexpr uint8_t IRQ_YIELD         = 0xF1;
+#if ARCH(i686)
-	constexpr uint8_t IRQ_IPI           = 0xF2;
+	constexpr uint8_t IRQ_SYSCALL       = 0xF0; // hard coded in kernel/API/Syscall.h
-	constexpr uint8_t IRQ_TIMER         = 0xF3;
+#endif
+	constexpr uint8_t IRQ_IPI           = 0xF1;
+	constexpr uint8_t IRQ_TIMER         = 0xF2;
 
 #if ARCH(x86_64)
 	struct GateDescriptor

kernel/include/kernel/Input/InputDevice.h

@@ -15,6 +15,7 @@ namespace Kernel
 		{
 			Mouse,
 			Keyboard,
+			Joystick,
 		};
 
 	public:

kernel/include/kernel/Input/PS2/Controller.h

@@ -14,6 +14,15 @@ namespace Kernel::Input
 
 	class PS2Controller
 	{
+	public:
+		enum class DeviceType
+		{
+			None,
+			Unknown,
+			Keyboard,
+			Mouse,
+		};
+
 	public:
 		static BAN::ErrorOr<void> initialize(uint8_t scancode_set);
 		static PS2Controller& get();
@@ -24,10 +33,12 @@ namespace Kernel::Input
 		// Returns true, if byte is used as command, if returns false, byte is meant to device
 		bool handle_command_byte(PS2Device*, uint8_t);
 
+		uint8_t data_port() const { return m_data_port; }
+
 	private:
 		PS2Controller() = default;
 		BAN::ErrorOr<void> initialize_impl(uint8_t scancode_set);
-		BAN::ErrorOr<void> identify_device(uint8_t, uint8_t scancode_set);
+		BAN::ErrorOr<DeviceType> identify_device(uint8_t);
 
 		void device_initialize_task(void*);
 
@@ -61,6 +72,9 @@ namespace Kernel::Input
 		};
 
 	private:
+		uint16_t m_command_port { PS2::IOPort::COMMAND };
+		uint16_t m_data_port    { PS2::IOPort::DATA };
+
 		BAN::RefPtr<PS2Device> m_devices[2];
 
 		Mutex m_mutex;

kernel/include/kernel/InterruptStack.h

@@ -27,7 +27,6 @@ namespace Kernel
 		uintptr_t r10;
 		uintptr_t r9;
 		uintptr_t r8;
-
 		uintptr_t rdi;
 		uintptr_t rsi;
 		uintptr_t rbp;
@@ -36,6 +35,18 @@ namespace Kernel
 		uintptr_t rcx;
 		uintptr_t rax;
 	};
+	struct YieldRegisters
+	{
+		uintptr_t r15;
+		uintptr_t r14;
+		uintptr_t r13;
+		uintptr_t r12;
+		uintptr_t rbp;
+		uintptr_t rbx;
+		uintptr_t ret;
+		uintptr_t sp;
+		uintptr_t ip;
+	};
 #elif ARCH(i686)
 	struct InterruptRegisters
 	{
@@ -48,6 +59,16 @@ namespace Kernel
 		uintptr_t ecx;
 		uintptr_t eax;
 	};
+	struct YieldRegisters
+	{
+		uintptr_t ebp;
+		uintptr_t edi;
+		uintptr_t esi;
+		uintptr_t ebx;
+		uintptr_t ret;
+		uintptr_t sp;
+		uintptr_t ip;
+	};
 #endif
 
 }

kernel/include/kernel/Lock/RWLock.h

@@ -0,0 +1,97 @@
+#pragma once
+
+#include <kernel/Lock/Mutex.h>
+#include <kernel/Lock/LockGuard.h>
+
+namespace Kernel
+{
+
+	class RWLock
+	{
+		BAN_NON_COPYABLE(RWLock);
+		BAN_NON_MOVABLE(RWLock);
+	public:
+		RWLock() = default;
+
+		void rd_lock()
+		{
+			LockGuard _(m_mutex);
+			while (m_writers_waiting > 0 || m_writer_active)
+				m_thread_blocker.block_indefinite(&m_mutex);
+			m_readers_active++;
+		}
+
+		void rd_unlock()
+		{
+			LockGuard _(m_mutex);
+			if (--m_readers_active == 0)
+				m_thread_blocker.unblock();
+		}
+
+		void wr_lock()
+		{
+			LockGuard _(m_mutex);
+			m_writers_waiting++;
+			while (m_readers_active > 0 || m_writer_active)
+				m_thread_blocker.block_indefinite(&m_mutex);
+			m_writers_waiting--;
+			m_writer_active = true;
+		}
+
+
+		void wr_unlock()
+		{
+			LockGuard _(m_mutex);
+			m_writer_active = false;
+			m_thread_blocker.unblock();
+		}
+
+	private:
+		Mutex m_mutex;
+		ThreadBlocker m_thread_blocker;
+		uint32_t m_readers_active  { 0 };
+		uint32_t m_writers_waiting { 0 };
+		bool m_writer_active { false };
+	};
+
+	class RWLockRDGuard
+	{
+		BAN_NON_COPYABLE(RWLockRDGuard);
+		BAN_NON_MOVABLE(RWLockRDGuard);
+	public:
+		RWLockRDGuard(RWLock& lock)
+			: m_lock(lock)
+		{
+			m_lock.rd_lock();
+		}
+
+		~RWLockRDGuard()
+		{
+			m_lock.rd_unlock();
+		}
+
+	private:
+		RWLock& m_lock;
+	};
+
+	class RWLockWRGuard
+	{
+		BAN_NON_COPYABLE(RWLockWRGuard);
+		BAN_NON_MOVABLE(RWLockWRGuard);
+	public:
+		RWLockWRGuard(RWLock& lock)
+			: m_lock(lock)
+		{
+			m_lock.wr_lock();
+		}
+
+		~RWLockWRGuard()
+		{
+			m_lock.wr_unlock();
+		}
+
+	private:
+		RWLock& m_lock;
+	};
+
+}

kernel/include/kernel/Memory/ByteRingBuffer.h

@@ -0,0 +1,76 @@
+#pragma once
+
+#include <BAN/ByteSpan.h>
+#include <BAN/UniqPtr.h>
+#include <BAN/Vector.h>
+
+#include <kernel/Memory/Types.h>
+
+namespace Kernel
+{
+
+	class ByteRingBuffer
+	{
+	public:
+		static BAN::ErrorOr<BAN::UniqPtr<ByteRingBuffer>> create(size_t size);
+		~ByteRingBuffer();
+
+		void push(BAN::ConstByteSpan data)
+		{
+			ASSERT(data.size() + m_size <= m_capacity);
+			uint8_t* buffer_head = reinterpret_cast<uint8_t*>(m_vaddr) + (m_tail + m_size) % m_capacity;
+			memcpy(buffer_head, data.data(), data.size());
+			m_size += data.size();
+		}
+
+		void pop(size_t size)
+		{
+			ASSERT(size <= m_size);
+			m_tail = (m_tail + size) % m_capacity;
+			m_size -= size;
+		}
+
+		void pop_back(size_t size)
+		{
+			ASSERT(size <= m_size);
+			m_size -= size;
+		}
+
+		BAN::ConstByteSpan get_data() const
+		{
+			const uint8_t* base = reinterpret_cast<const uint8_t*>(m_vaddr);
+			return { base + m_tail, m_size };
+		}
+
+		uint8_t front() const
+		{
+			ASSERT(!empty());
+			return reinterpret_cast<const uint8_t*>(m_vaddr)[m_tail];
+		}
+
+		uint8_t back() const
+		{
+			ASSERT(!empty());
+			return reinterpret_cast<const uint8_t*>(m_vaddr)[m_tail + m_size];
+		}
+
+		bool empty() const { return m_size == 0; }
+		bool full() const { return m_size == m_capacity; }
+		size_t free() const { return m_capacity - m_size; }
+		size_t size() const { return m_size; }
+		size_t capacity() const { return m_capacity; }
+
+	private:
+		ByteRingBuffer(size_t capacity)
+			: m_capacity(capacity)
+		{ }
+
+	private:
+		size_t m_size { 0 };
+		size_t m_tail { 0 };
+		const size_t m_capacity;
+
+		vaddr_t m_vaddr { 0 };
+	};
+
+}

kernel/include/kernel/Memory/FileBackedRegion.h

@@ -27,18 +27,19 @@ namespace Kernel
 		BAN_NON_MOVABLE(FileBackedRegion);
 
 	public:
-		static BAN::ErrorOr<BAN::UniqPtr<FileBackedRegion>> create(BAN::RefPtr<Inode>, PageTable&, off_t offset, size_t size, AddressRange address_range, Type, PageTable::flags_t);
+		static BAN::ErrorOr<BAN::UniqPtr<FileBackedRegion>> create(BAN::RefPtr<Inode>, PageTable&, off_t offset, size_t size, AddressRange address_range, Type, PageTable::flags_t, int status_flags);
 		~FileBackedRegion();
 
-		virtual BAN::ErrorOr<void> msync(vaddr_t, size_t, int) override;
+		BAN::ErrorOr<void> msync(vaddr_t, size_t, int) override;
 
-		virtual BAN::ErrorOr<BAN::UniqPtr<MemoryRegion>> clone(PageTable& new_page_table) override;
+		BAN::ErrorOr<BAN::UniqPtr<MemoryRegion>> clone(PageTable& new_page_table) override;
+		BAN::ErrorOr<BAN::UniqPtr<MemoryRegion>> split(size_t offset) override;
 
 	protected:
-		virtual BAN::ErrorOr<bool> allocate_page_containing_impl(vaddr_t vaddr, bool wants_write) override;
+		BAN::ErrorOr<bool> allocate_page_containing_impl(vaddr_t vaddr, bool wants_write) override;
 
 	private:
-		FileBackedRegion(BAN::RefPtr<Inode>, PageTable&, off_t offset, ssize_t size, Type flags, PageTable::flags_t page_flags);
+		FileBackedRegion(BAN::RefPtr<Inode>, PageTable&, off_t offset, ssize_t size, Type type, PageTable::flags_t flags, int status_flags);
 
 	private:
 		BAN::RefPtr<Inode> m_inode;

kernel/include/kernel/Memory/Heap.h

@@ -18,6 +18,8 @@ namespace Kernel
 		static void initialize();
 		static Heap& get();
 
+		void release_boot_modules();
+
 		paddr_t take_free_page();
 		void release_page(paddr_t);
 

kernel/include/kernel/Memory/MemoryBackedRegion.h

@@ -11,22 +11,37 @@ namespace Kernel
 		BAN_NON_MOVABLE(MemoryBackedRegion);
 
 	public:
-		static BAN::ErrorOr<BAN::UniqPtr<MemoryBackedRegion>> create(PageTable&, size_t size, AddressRange, Type, PageTable::flags_t);
+		static BAN::ErrorOr<BAN::UniqPtr<MemoryBackedRegion>> create(PageTable&, size_t size, AddressRange, Type, PageTable::flags_t, int status_flags);
 		~MemoryBackedRegion();
 
-		virtual BAN::ErrorOr<BAN::UniqPtr<MemoryRegion>> clone(PageTable& new_page_table) override;
+		BAN::ErrorOr<BAN::UniqPtr<MemoryRegion>> clone(PageTable& new_page_table) override;
+		BAN::ErrorOr<BAN::UniqPtr<MemoryRegion>> split(size_t offset) override;
 
-		virtual BAN::ErrorOr<void> msync(vaddr_t, size_t, int) override { return {}; }
+		BAN::ErrorOr<void> msync(vaddr_t, size_t, int) override { return {}; }
 
 		// Copy data from buffer into this region
 		// This can fail if no memory is mapped and no free memory was available
 		BAN::ErrorOr<void> copy_data_to_region(size_t offset_into_region, const uint8_t* buffer, size_t buffer_size);
 
 	protected:
-		virtual BAN::ErrorOr<bool> allocate_page_containing_impl(vaddr_t vaddr, bool wants_write) override;
+		BAN::ErrorOr<bool> allocate_page_containing_impl(vaddr_t vaddr, bool wants_write) override;
 
 	private:
-		MemoryBackedRegion(PageTable&, size_t size, Type, PageTable::flags_t);
+		MemoryBackedRegion(PageTable&, size_t size, Type, PageTable::flags_t, int status_flags);
+
+	private:
+		struct PhysicalPage
+		{
+			PhysicalPage(paddr_t paddr)
+				: paddr(paddr)
+			{ }
+			~PhysicalPage();
+
+			BAN::Atomic<uint32_t> ref_count { 1 };
+			const paddr_t paddr;
+		};
+		BAN::Vector<PhysicalPage*> m_physical_pages;
+		Mutex m_mutex;
 	};
 
 }

kernel/include/kernel/Memory/MemoryRegion.h

@@ -33,12 +33,17 @@ namespace Kernel
 		bool contains(vaddr_t address) const;
 		bool contains_fully(vaddr_t address, size_t size) const;
 		bool overlaps(vaddr_t address, size_t size) const;
+		bool is_contained_by(vaddr_t address, size_t size) const;
 
 		bool writable() const { return m_flags & PageTable::Flags::ReadWrite; }
 
 		size_t size() const { return m_size; }
 		vaddr_t vaddr() const { return m_vaddr; }
 
+		int status_flags() const { return m_status_flags; }
+		Type type() const { return m_type; }
+		PageTable::flags_t flags() const { return m_flags; }
+
 		size_t virtual_page_count() const { return BAN::Math::div_round_up<size_t>(m_size, PAGE_SIZE); }
 		size_t physical_page_count() const { return m_physical_page_count; }
 
@@ -46,6 +51,7 @@ namespace Kernel
 		void unpin();
 		void wait_not_pinned();
 
+		BAN::ErrorOr<void> mprotect(PageTable::flags_t);
 		virtual BAN::ErrorOr<void> msync(vaddr_t, size_t, int) = 0;
 
 		// Returns error if no memory was available
@@ -54,18 +60,20 @@ namespace Kernel
 		BAN::ErrorOr<bool> allocate_page_containing(vaddr_t address, bool wants_write);
 
 		virtual BAN::ErrorOr<BAN::UniqPtr<MemoryRegion>> clone(PageTable& new_page_table) = 0;
+		virtual BAN::ErrorOr<BAN::UniqPtr<MemoryRegion>> split(size_t offset) = 0;
 
 	protected:
-		MemoryRegion(PageTable&, size_t size, Type type, PageTable::flags_t flags);
+		MemoryRegion(PageTable&, size_t size, Type type, PageTable::flags_t flags, int status_flags);
 		BAN::ErrorOr<void> initialize(AddressRange);
 
 		virtual BAN::ErrorOr<bool> allocate_page_containing_impl(vaddr_t address, bool wants_write) = 0;
 
 	protected:
 		PageTable& m_page_table;
-		const size_t m_size;
+		size_t m_size { 0 };
 		const Type m_type;
-		const PageTable::flags_t m_flags;
+		PageTable::flags_t m_flags;
+		const int m_status_flags;
 		vaddr_t m_vaddr { 0 };
 		size_t m_physical_page_count { 0 };
 

kernel/include/kernel/Memory/PageTable.h

@@ -100,19 +100,21 @@ namespace Kernel
 		static BAN::ErrorOr<PageTable*> create_userspace();
 		~PageTable();
 
-		void unmap_page(vaddr_t, bool send_smp_message = true);
+		void unmap_page(vaddr_t, bool invalidate = true);
 		void unmap_range(vaddr_t, size_t bytes);
 
-		void map_page_at(paddr_t, vaddr_t, flags_t, MemoryType = MemoryType::Normal, bool send_smp_message = true);
+		void map_page_at(paddr_t, vaddr_t, flags_t, MemoryType = MemoryType::Normal, bool invalidate = true);
 		void map_range_at(paddr_t, vaddr_t, size_t bytes, flags_t, MemoryType = MemoryType::Normal);
 
+		void remove_writable_from_range(vaddr_t, size_t);
+
 		paddr_t physical_address_of(vaddr_t) const;
 		flags_t get_page_flags(vaddr_t) const;
 
 		bool is_page_free(vaddr_t) const;
 		bool is_range_free(vaddr_t, size_t bytes) const;
 
-		bool reserve_page(vaddr_t, bool only_free = true);
+		bool reserve_page(vaddr_t, bool only_free = true, bool invalidate = true);
 		bool reserve_range(vaddr_t, size_t bytes, bool only_free = true);
 
 		vaddr_t reserve_free_page(vaddr_t first_address, vaddr_t last_address = UINTPTR_MAX);
@@ -121,6 +123,9 @@ namespace Kernel
 		void load();
 		void initial_load();
 
+		void invalidate_page(vaddr_t addr, bool send_smp_message) { invalidate_range(addr, 1, send_smp_message); }
+		void invalidate_range(vaddr_t addr, size_t pages, bool send_smp_message);
+
 		InterruptState lock() const { return m_lock.lock(); }
 		void unlock(InterruptState state) const { m_lock.unlock(state); }
 
@@ -133,8 +138,6 @@ namespace Kernel
 		void map_kernel_memory();
 		void prepare_fast_page();
 
-		static void invalidate(vaddr_t, bool send_smp_message);
-
 		static void map_fast_page(paddr_t);
 		static void unmap_fast_page();
 

kernel/include/kernel/Memory/PhysicalRange.h

@@ -10,7 +10,7 @@ namespace Kernel
 	class PhysicalRange
 	{
 	public:
-		PhysicalRange(paddr_t, size_t);
+		PhysicalRange(paddr_t, uint64_t);
 
 		paddr_t reserve_page();
 		void release_page(paddr_t);

kernel/include/kernel/Memory/SharedMemoryObject.h

@@ -6,6 +6,8 @@
 #include <kernel/Lock/SpinLock.h>
 #include <kernel/Memory/MemoryRegion.h>
 
+#include <fcntl.h>
+
 namespace Kernel
 {
 
@@ -55,15 +57,17 @@ namespace Kernel
 	public:
 		static BAN::ErrorOr<BAN::UniqPtr<SharedMemoryObject>> create(BAN::RefPtr<SharedMemoryObjectManager::Object>, PageTable&, AddressRange);
 
-		virtual BAN::ErrorOr<BAN::UniqPtr<MemoryRegion>> clone(PageTable& new_page_table) override;
+		BAN::ErrorOr<BAN::UniqPtr<MemoryRegion>> clone(PageTable& new_page_table) override;
-		virtual BAN::ErrorOr<void> msync(vaddr_t, size_t, int) override { return {}; }
+		BAN::ErrorOr<BAN::UniqPtr<MemoryRegion>> split(size_t offset) override;
+
+		BAN::ErrorOr<void> msync(vaddr_t, size_t, int) override { return {}; }
 
 	protected:
-		virtual BAN::ErrorOr<bool> allocate_page_containing_impl(vaddr_t vaddr, bool wants_write) override;
+		BAN::ErrorOr<bool> allocate_page_containing_impl(vaddr_t vaddr, bool wants_write) override;
 
 	private:
 		SharedMemoryObject(BAN::RefPtr<SharedMemoryObjectManager::Object> object, PageTable& page_table)
-			: MemoryRegion(page_table, object->size, MemoryRegion::Type::SHARED, object->flags)
+			: MemoryRegion(page_table, object->size, MemoryRegion::Type::SHARED, object->flags, O_EXEC | O_RDWR)
 			, m_object(object)
 		{ }
 

kernel/include/kernel/Memory/Types.h

@@ -4,7 +4,7 @@
 
 #if ARCH(x86_64)
 	#define KERNEL_OFFSET 0xFFFFFFFF80000000
-	#define USERSPACE_END 0xFFFF800000000000
+	#define USERSPACE_END 0x800000000000
 #elif ARCH(i686)
 	#define KERNEL_OFFSET 0xC0000000
 	#define USERSPACE_END 0xC0000000

kernel/include/kernel/Memory/VirtualRange.h

@@ -20,8 +20,6 @@ namespace Kernel
 		static BAN::ErrorOr<BAN::UniqPtr<VirtualRange>> create_to_vaddr_range(PageTable&, vaddr_t vaddr_start, vaddr_t vaddr_end, size_t, PageTable::flags_t flags, bool preallocate_pages, bool add_guard_pages);
 		~VirtualRange();
 
-		BAN::ErrorOr<BAN::UniqPtr<VirtualRange>> clone(PageTable&);
-
 		vaddr_t vaddr() const { return m_vaddr + (m_has_guard_pages ? PAGE_SIZE : 0); }
 		size_t size() const { return m_size - (m_has_guard_pages ? 2 * PAGE_SIZE : 0); }
 		PageTable::flags_t flags() const { return m_flags; }
@@ -38,7 +36,7 @@ namespace Kernel
 
 		bool contains(vaddr_t address) const { return vaddr() <= address && address < vaddr() + size(); }
 
-		BAN::ErrorOr<void> allocate_page_for_demand_paging(vaddr_t address);
+		BAN::ErrorOr<bool> allocate_page_for_demand_paging(vaddr_t address);
 
 	private:
 		VirtualRange(PageTable&, bool preallocated, bool has_guard_pages, vaddr_t, size_t, PageTable::flags_t);

kernel/include/kernel/Networking/ARPTable.h

@@ -31,35 +31,18 @@ namespace Kernel
 
 	public:
 		static BAN::ErrorOr<BAN::UniqPtr<ARPTable>> create();
-		~ARPTable();
 
 		BAN::ErrorOr<BAN::MACAddress> get_mac_from_ipv4(NetworkInterface&, BAN::IPv4Address);
 
-		void add_arp_packet(NetworkInterface&, BAN::ConstByteSpan);
+		BAN::ErrorOr<void> handle_arp_packet(NetworkInterface&, BAN::ConstByteSpan);
 
 	private:
-		ARPTable();
+		ARPTable() = default;
-
-		void packet_handle_task();
-		BAN::ErrorOr<void> handle_arp_packet(NetworkInterface&, const ARPPacket&);
 
 	private:
-		struct PendingArpPacket
+		SpinLock m_arp_table_lock;
-		{
-			NetworkInterface& interface;
-			ARPPacket packet;
-		};
-
-	private:
-		SpinLock m_table_lock;
-		SpinLock m_pending_lock;
-
 		BAN::HashMap<BAN::IPv4Address, BAN::MACAddress> m_arp_table;
 
-		Thread*										m_thread { nullptr };
-		BAN::CircularQueue<PendingArpPacket, 128>	m_pending_packets;
-		ThreadBlocker								m_pending_thread_blocker;
-
 		friend class BAN::UniqPtr<ARPTable>;
 	};
 

kernel/include/kernel/Networking/E1000/E1000.h

@@ -23,14 +23,14 @@ namespace Kernel
 		static BAN::ErrorOr<BAN::RefPtr<E1000>> create(PCI::Device&);
 		~E1000();
 
-		virtual BAN::MACAddress get_mac_address() const override { return m_mac_address; }
+		BAN::MACAddress get_mac_address() const override { return m_mac_address; }
 
-		virtual bool link_up() override { return m_link_up; }
+		bool link_up() override { return m_link_up; }
-		virtual int link_speed() override;
+		int link_speed() override;
 
-		virtual size_t payload_mtu() const override { return E1000_RX_BUFFER_SIZE - sizeof(EthernetHeader); }
+		size_t payload_mtu() const override { return E1000_RX_BUFFER_SIZE - sizeof(EthernetHeader); }
 
-		virtual void handle_irq() final override;
+		void handle_irq() final override;
 
 	protected:
 		E1000(PCI::Device& pci_device)
@@ -45,12 +45,12 @@ namespace Kernel
 		uint32_t read32(uint16_t reg);
 		void write32(uint16_t reg, uint32_t value);
 
-		virtual BAN::ErrorOr<void> send_bytes(BAN::MACAddress destination, EtherType protocol, BAN::ConstByteSpan) override;
+		BAN::ErrorOr<void> send_bytes(BAN::MACAddress destination, EtherType protocol, BAN::Span<const BAN::ConstByteSpan> payload) override;
 
-		virtual bool can_read_impl() const override { return false; }
+		bool can_read_impl() const override { return false; }
-		virtual bool can_write_impl() const override { return false; }
+		bool can_write_impl() const override { return false; }
-		virtual bool has_error_impl() const override { return false; }
+		bool has_error_impl() const override { return false; }
-		virtual bool has_hungup_impl() const override { return false; }
+		bool has_hungup_impl() const override { return false; }
 
 	private:
 		BAN::ErrorOr<void> read_mac_address();
@@ -61,7 +61,7 @@ namespace Kernel
 		void enable_link();
 		BAN::ErrorOr<void> enable_interrupt();
 
-		void handle_receive();
+		void receive_thread();
 
 	protected:
 		PCI::Device&					m_pci_device;
@@ -75,6 +75,10 @@ namespace Kernel
 		BAN::UniqPtr<DMARegion>	m_tx_descriptor_region;
 		SpinLock m_lock;
 
+		bool m_thread_should_die { false };
+		BAN::Atomic<bool> m_thread_is_dead { true };
+		ThreadBlocker m_thread_blocker;
+
 		BAN::MACAddress	m_mac_address {};
 		bool			m_link_up { false };
 

kernel/include/kernel/Networking/E1000/E1000E.h

@@ -12,8 +12,8 @@ namespace Kernel
 		static BAN::ErrorOr<BAN::RefPtr<E1000E>> create(PCI::Device&);
 
 	protected:
-		virtual void detect_eeprom() override;
+		void detect_eeprom() override;
-		virtual uint32_t eeprom_read(uint8_t addr) override;
+		uint32_t eeprom_read(uint8_t addr) override;
 
 	private:
 		E1000E(PCI::Device& pci_device)

kernel/include/kernel/Networking/IPv4Layer.h

@@ -38,14 +38,13 @@ namespace Kernel
 
 	public:
 		static BAN::ErrorOr<BAN::UniqPtr<IPv4Layer>> create();
-		~IPv4Layer();
 
 		ARPTable& arp_table() { return *m_arp_table; }
 
-		void add_ipv4_packet(NetworkInterface&, BAN::ConstByteSpan);
+		BAN::ErrorOr<void> handle_ipv4_packet(NetworkInterface&, BAN::ConstByteSpan);
 
 		virtual void unbind_socket(uint16_t port) override;
-		virtual BAN::ErrorOr<void> bind_socket_to_unused(BAN::RefPtr<NetworkSocket>, const sockaddr* send_address, socklen_t send_address_len) override;
+		virtual BAN::ErrorOr<void> bind_socket_with_target(BAN::RefPtr<NetworkSocket>, const sockaddr* target_address, socklen_t target_address_len) override;
 		virtual BAN::ErrorOr<void> bind_socket_to_address(BAN::RefPtr<NetworkSocket>, const sockaddr* address, socklen_t address_len) override;
 		virtual BAN::ErrorOr<void> get_socket_address(BAN::RefPtr<NetworkSocket>, sockaddr* address, socklen_t* address_len) override;
 
@@ -55,33 +54,15 @@ namespace Kernel
 		virtual size_t header_size() const override { return sizeof(IPv4Header); }
 
 	private:
-		IPv4Layer();
+		IPv4Layer() = default;
 
-		void add_ipv4_header(BAN::ByteSpan packet, BAN::IPv4Address src_ipv4, BAN::IPv4Address dst_ipv4, uint8_t protocol) const;
+		BAN::ErrorOr<in_port_t> find_free_port();
-
-		void packet_handle_task();
-		BAN::ErrorOr<void> handle_ipv4_packet(NetworkInterface&, BAN::ByteSpan);
 
 	private:
-		struct PendingIPv4Packet
+		BAN::UniqPtr<ARPTable> m_arp_table;
-		{
-			NetworkInterface& interface;
-		};
 
-	private:
+		RecursiveSpinLock m_bound_socket_lock;
-		RecursiveSpinLock		m_bound_socket_lock;
+		BAN::HashMap<int, BAN::WeakPtr<NetworkSocket>> m_bound_sockets;
-
-		BAN::UniqPtr<ARPTable>	m_arp_table;
-		Thread*					m_thread { nullptr };
-
-		static constexpr size_t pending_packet_buffer_size = 128 * PAGE_SIZE;
-		BAN::UniqPtr<VirtualRange>					m_pending_packet_buffer;
-		BAN::CircularQueue<PendingIPv4Packet, 128>	m_pending_packets;
-		ThreadBlocker									m_pending_thread_blocker;
-		SpinLock									m_pending_lock;
-		size_t										m_pending_total_size { 0 };
-
-		BAN::HashMap<int, BAN::WeakPtr<NetworkSocket>>	m_bound_sockets;
 
 		friend class BAN::UniqPtr<IPv4Layer>;
 	};

kernel/include/kernel/Networking/Loopback.h

@@ -9,6 +9,7 @@ namespace Kernel
 	{
 	public:
 		static constexpr size_t buffer_size = BAN::numeric_limits<uint16_t>::max() + 1;
+		static constexpr size_t buffer_count = 32;
 
 	public:
 		static BAN::ErrorOr<BAN::RefPtr<LoopbackInterface>> create();
@@ -24,8 +25,9 @@ namespace Kernel
 		LoopbackInterface()
 			: NetworkInterface(Type::Loopback)
 		{}
+		~LoopbackInterface();
 
-		BAN::ErrorOr<void> send_bytes(BAN::MACAddress destination, EtherType protocol, BAN::ConstByteSpan) override;
+		BAN::ErrorOr<void> send_bytes(BAN::MACAddress destination, EtherType protocol, BAN::Span<const BAN::ConstByteSpan> payload) override;
 
 		bool can_read_impl() const override { return false; }
 		bool can_write_impl() const override { return false; }
@@ -33,8 +35,27 @@ namespace Kernel
 		bool has_hungup_impl() const override { return false; }
 
 	private:
-		SpinLock m_buffer_lock;
+		void receive_thread();
+
+	private:
+		struct Descriptor
+		{
+			uint8_t* addr;
+			uint32_t size;
+			uint8_t state;
+		};
+
+	private:
+		Mutex m_buffer_lock;
 		BAN::UniqPtr<VirtualRange> m_buffer;
+
+		uint32_t m_buffer_tail { 0 };
+		uint32_t m_buffer_head { 0 };
+		Descriptor m_descriptors[buffer_count] {};
+
+		bool m_thread_should_die { false };
+		BAN::Atomic<bool> m_thread_is_dead { true };
+		ThreadBlocker m_thread_blocker;
 	};
 
 }

kernel/include/kernel/Networking/NetworkInterface.h

@@ -60,7 +60,11 @@ namespace Kernel
 		virtual dev_t rdev() const override { return m_rdev; }
 		virtual BAN::StringView name() const override { return m_name; }
 
-		virtual BAN::ErrorOr<void> send_bytes(BAN::MACAddress destination, EtherType protocol, BAN::ConstByteSpan) = 0;
+		BAN::ErrorOr<void> send_bytes(BAN::MACAddress destination, EtherType protocol, BAN::ConstByteSpan payload)
+		{
+			return send_bytes(destination, protocol, { &payload, 1 });
+		}
+		virtual BAN::ErrorOr<void> send_bytes(BAN::MACAddress destination, EtherType protocol, BAN::Span<const BAN::ConstByteSpan> payload) = 0;
 
 	private:
 		const Type m_type;

kernel/include/kernel/Networking/NetworkLayer.h

@@ -11,7 +11,7 @@ namespace Kernel
 		BAN::IPv4Address src_ipv4				{ 0 };
 		BAN::IPv4Address dst_ipv4				{ 0 };
 		BAN::NetworkEndian<uint16_t> protocol	{ 0 };
-		BAN::NetworkEndian<uint16_t> extra		{ 0 };
+		BAN::NetworkEndian<uint16_t> length		{ 0 };
 	};
 	static_assert(sizeof(PseudoHeader) == 12);
 
@@ -23,7 +23,7 @@ namespace Kernel
 		virtual ~NetworkLayer() {}
 
 		virtual void unbind_socket(uint16_t port) = 0;
-		virtual BAN::ErrorOr<void> bind_socket_to_unused(BAN::RefPtr<NetworkSocket>, const sockaddr* send_address, socklen_t send_address_len) = 0;
+		virtual BAN::ErrorOr<void> bind_socket_with_target(BAN::RefPtr<NetworkSocket>, const sockaddr* target_address, socklen_t target_address_len) = 0;
 		virtual BAN::ErrorOr<void> bind_socket_to_address(BAN::RefPtr<NetworkSocket>, const sockaddr* address, socklen_t address_len) = 0;
 		virtual BAN::ErrorOr<void> get_socket_address(BAN::RefPtr<NetworkSocket>, sockaddr* address, socklen_t* address_len) = 0;
 
@@ -36,6 +36,7 @@ namespace Kernel
 		NetworkLayer() = default;
 	};
 
-	uint16_t calculate_internet_checksum(BAN::ConstByteSpan packet, const PseudoHeader& pseudo_header);
+	uint16_t calculate_internet_checksum(BAN::ConstByteSpan buffer);
+	uint16_t calculate_internet_checksum(BAN::Span<const BAN::ConstByteSpan> buffers);
 
 }

kernel/include/kernel/Networking/NetworkSocket.h

@@ -5,6 +5,8 @@
 #include <kernel/Networking/NetworkInterface.h>
 #include <kernel/Networking/NetworkLayer.h>
 
+#include <netinet/in.h>
+
 namespace Kernel
 {
 
@@ -24,18 +26,30 @@ namespace Kernel
 		static constexpr uint16_t PORT_NONE = 0;
 
 	public:
-		void bind_interface_and_port(NetworkInterface*, uint16_t port);
+		void bind_address_and_port(const sockaddr*, socklen_t);
 		~NetworkSocket();
 
-		NetworkInterface& interface() { ASSERT(m_interface); return *m_interface; }
+		BAN::ErrorOr<BAN::RefPtr<NetworkInterface>> interface(const sockaddr* target, socklen_t target_len);
 
 		virtual size_t protocol_header_size() const = 0;
-		virtual void add_protocol_header(BAN::ByteSpan packet, uint16_t dst_port, PseudoHeader) = 0;
+		virtual void get_protocol_header(BAN::ByteSpan header, BAN::ConstByteSpan payload, uint16_t dst_port, PseudoHeader) = 0;
 		virtual NetworkProtocol protocol() const = 0;
 
 		virtual void receive_packet(BAN::ConstByteSpan, const sockaddr* sender, socklen_t sender_len) = 0;
 
-		bool is_bound() const { return m_interface != nullptr; }
+		bool is_bound() const { return m_address_len >= static_cast<socklen_t>(sizeof(sa_family_t)) && m_address.ss_family != AF_UNSPEC; }
+		in_port_t bound_port() const
+		{
+			ASSERT(is_bound());
+			ASSERT(m_address.ss_family == AF_INET && m_address_len >= static_cast<socklen_t>(sizeof(sockaddr_in)));
+			return BAN::network_endian_to_host(reinterpret_cast<const sockaddr_in*>(&m_address)->sin_port);
+		}
+
+		const sockaddr* address() const { return reinterpret_cast<const sockaddr*>(&m_address); }
+		socklen_t address_len() const { return m_address_len; }
+
+	private:
+		bool can_interface_send_to(const NetworkInterface&, const sockaddr*, socklen_t) const;
 
 	protected:
 		NetworkSocket(NetworkLayer&, const Socket::Info&);
@@ -45,9 +59,9 @@ namespace Kernel
 		virtual BAN::ErrorOr<void> getpeername_impl(sockaddr*, socklen_t*) override = 0;
 
 	protected:
-		NetworkLayer&		m_network_layer;
+		NetworkLayer&    m_network_layer;
-		NetworkInterface*	m_interface	= nullptr;
+		sockaddr_storage m_address       { .ss_family = AF_UNSPEC, .ss_storage = {} };
-		uint16_t			m_port { PORT_NONE };
+		socklen_t        m_address_len   { 0 };
 	};
 
 }

kernel/include/kernel/Networking/RTL8169/RTL8169.h

@@ -29,9 +29,11 @@ namespace Kernel
 			: NetworkInterface(Type::Ethernet)
 			, m_pci_device(pci_device)
 		{ }
+		~RTL8169();
+
 		BAN::ErrorOr<void> initialize();
 
-		virtual BAN::ErrorOr<void> send_bytes(BAN::MACAddress destination, EtherType protocol, BAN::ConstByteSpan) override;
+		virtual BAN::ErrorOr<void> send_bytes(BAN::MACAddress destination, EtherType protocol, BAN::Span<const BAN::ConstByteSpan>) override;
 
 		virtual bool can_read_impl() const override { return false; }
 		virtual bool can_write_impl() const override { return false; }
@@ -47,7 +49,7 @@ namespace Kernel
 		void enable_link();
 		BAN::ErrorOr<void> enable_interrupt();
 
-		void handle_receive();
+		void receive_thread();
 
 	protected:
 		PCI::Device&					m_pci_device;
@@ -63,6 +65,9 @@ namespace Kernel
 		BAN::UniqPtr<DMARegion>	m_tx_descriptor_region;
 
 		SpinLock m_lock;
+
+		bool m_thread_should_die { false };
+		BAN::Atomic<bool> m_thread_is_dead { true };
 		ThreadBlocker m_thread_blocker;
 
 		uint32_t m_rx_current { 0 };

kernel/include/kernel/Networking/TCPSocket.h

@@ -1,9 +1,10 @@
 #pragma once
 
+#include <BAN/HashMap.h>
 #include <BAN/Endianness.h>
 #include <BAN/Queue.h>
 #include <kernel/Lock/Mutex.h>
-#include <kernel/Memory/VirtualRange.h>
+#include <kernel/Memory/ByteRingBuffer.h>
 #include <kernel/Networking/NetworkInterface.h>
 #include <kernel/Networking/NetworkSocket.h>
 #include <kernel/Thread.h>
@@ -49,26 +50,30 @@ namespace Kernel
 		static BAN::ErrorOr<BAN::RefPtr<TCPSocket>> create(NetworkLayer&, const Info&);
 		~TCPSocket();
 
-		virtual NetworkProtocol protocol() const override { return NetworkProtocol::TCP; }
+		NetworkProtocol protocol() const override { return NetworkProtocol::TCP; }
 
-		virtual size_t protocol_header_size() const override { return sizeof(TCPHeader) + m_tcp_options_bytes; }
+		size_t protocol_header_size() const override { return sizeof(TCPHeader) + m_tcp_options_bytes; }
-		virtual void add_protocol_header(BAN::ByteSpan packet, uint16_t dst_port, PseudoHeader) override;
+		void get_protocol_header(BAN::ByteSpan header, BAN::ConstByteSpan payload, uint16_t dst_port, PseudoHeader) override;
 
 	protected:
-		virtual BAN::ErrorOr<long> accept_impl(sockaddr*, socklen_t*, int) override;
+		BAN::ErrorOr<long> accept_impl(sockaddr*, socklen_t*, int) override;
-		virtual BAN::ErrorOr<void> connect_impl(const sockaddr*, socklen_t) override;
+		BAN::ErrorOr<void> connect_impl(const sockaddr*, socklen_t) override;
-		virtual BAN::ErrorOr<void> listen_impl(int) override;
+		BAN::ErrorOr<void> listen_impl(int) override;
-		virtual BAN::ErrorOr<void> bind_impl(const sockaddr*, socklen_t) override;
+		BAN::ErrorOr<void> bind_impl(const sockaddr*, socklen_t) override;
-		virtual BAN::ErrorOr<size_t> sendto_impl(BAN::ConstByteSpan, const sockaddr*, socklen_t) override;
+		BAN::ErrorOr<size_t> recvmsg_impl(msghdr& message, int flags) override;
-		virtual BAN::ErrorOr<size_t> recvfrom_impl(BAN::ByteSpan, sockaddr*, socklen_t*) override;
+		BAN::ErrorOr<size_t> sendmsg_impl(const msghdr& message, int flags) override;
-		virtual BAN::ErrorOr<void> getpeername_impl(sockaddr*, socklen_t*) override;
+		BAN::ErrorOr<void> getpeername_impl(sockaddr*, socklen_t*) override;
+		BAN::ErrorOr<void> getsockopt_impl(int, int, void*, socklen_t*) override;
+		BAN::ErrorOr<void> setsockopt_impl(int, int, const void*, socklen_t) override;
 
-		virtual void receive_packet(BAN::ConstByteSpan, const sockaddr* sender, socklen_t sender_len) override;
+		BAN::ErrorOr<long> ioctl_impl(int, void*) override;
 
-		virtual bool can_read_impl() const override;
+		void receive_packet(BAN::ConstByteSpan, const sockaddr* sender, socklen_t sender_len) override;
-		virtual bool can_write_impl() const override;
+
-		virtual bool has_error_impl() const override { return false; }
+		bool can_read_impl() const override;
-		virtual bool has_hungup_impl() const override;
+		bool can_write_impl() const override;
+		bool has_error_impl() const override { return false; }
+		bool has_hungup_impl() const override;
 
 	private:
 		enum class State
@@ -88,33 +93,32 @@ namespace Kernel
 
 		struct RecvWindowInfo
 		{
-			uint32_t					start_seq	{ 0 }; // sequence number of first byte in buffer
+			uint32_t start_seq      { 0 }; // sequence number of first byte in buffer
 
-			bool						has_ghost_byte { false };
+			bool     has_ghost_byte { false };
 
-			uint32_t					data_size	{ 0 }; // number of bytes in this buffer
+			uint8_t  scale_shift    { 0 }; // window scale
-			uint8_t						scale_shift	{ 0 }; // window scale
+			BAN::UniqPtr<ByteRingBuffer> buffer;
-			BAN::UniqPtr<VirtualRange>	buffer;
 		};
 
 		struct SendWindowInfo
 		{
-			uint32_t					mss				{ 0 }; // maximum segment size
+			uint32_t mss             { 0 }; // maximum segment size
-			uint16_t					non_scaled_size	{ 0 }; // window size without scaling
+			uint16_t non_scaled_size { 0 }; // window size without scaling
-			uint8_t						scale_shift		{ 0 }; // window scale
+			uint8_t  scale_shift     { 0 }; // window scale
-			uint32_t 					scaled_size() const { return (uint32_t)non_scaled_size << scale_shift; }
+			uint32_t scaled_size() const { return (uint32_t)non_scaled_size << scale_shift; }
 
-			uint32_t					start_seq		{ 0 }; // sequence number of first byte in buffer
+			uint32_t start_seq       { 0 }; // sequence number of first byte in buffer
-			uint32_t					current_seq		{ 0 }; // sequence number of next send
+			uint32_t current_seq     { 0 }; // sequence number of next send
-			uint32_t					current_ack		{ 0 }; // sequence number aknowledged by connection
+			uint32_t current_ack     { 0 }; // sequence number aknowledged by connection
 
-			uint64_t					last_send_ms	{ 0 }; // last send time, used for retransmission timeout
+			uint64_t last_send_ms    { 0 }; // last send time, used for retransmission timeout
 
-			bool						has_ghost_byte { false };
+			bool     has_ghost_byte  { false };
+			bool     had_zero_window { false };
 
-			uint32_t					data_size		{ 0 }; // number of bytes in this buffer
+			uint32_t sent_size       { 0 }; // number of bytes in this buffer that have been sent
-			uint32_t					sent_size		{ 0 }; // number of bytes in this buffer that have been sent
+			BAN::UniqPtr<ByteRingBuffer> buffer;
-			BAN::UniqPtr<VirtualRange>	buffer;
 		};
 
 		struct ConnectionInfo
@@ -128,6 +132,8 @@ namespace Kernel
 		{
 			ConnectionInfo target;
 			uint32_t target_start_seq;
+			uint16_t maximum_seqment_size;
+			uint8_t window_scale;
 		};
 
 		struct ListenKey
@@ -162,8 +168,17 @@ namespace Kernel
 		State m_next_state		{ State::Closed };
 		uint8_t m_next_flags	{ 0 };
 
+		size_t m_last_sent_window_size { 0 };
+
 		Thread* m_thread { nullptr };
 
+		// TODO: actually support these
+		bool m_keep_alive { false };
+		bool m_no_delay { false };
+
+		bool m_should_send_zero_window { false };
+		bool m_should_send_window_update { false };
+
 		uint64_t m_time_wait_start_ms { 0 };
 
 		ThreadBlocker m_thread_blocker;

kernel/include/kernel/Networking/UDPSocket.h

@@ -25,23 +25,28 @@ namespace Kernel
 	public:
 		static BAN::ErrorOr<BAN::RefPtr<UDPSocket>> create(NetworkLayer&, const Socket::Info&);
 
-		virtual NetworkProtocol protocol() const override { return NetworkProtocol::UDP; }
+		NetworkProtocol protocol() const override { return NetworkProtocol::UDP; }
 
-		virtual size_t protocol_header_size() const override { return sizeof(UDPHeader); }
+		size_t protocol_header_size() const override { return sizeof(UDPHeader); }
-		virtual void add_protocol_header(BAN::ByteSpan packet, uint16_t dst_port, PseudoHeader) override;
+		void get_protocol_header(BAN::ByteSpan header, BAN::ConstByteSpan payload, uint16_t dst_port, PseudoHeader) override;
 
 	protected:
-		virtual void receive_packet(BAN::ConstByteSpan, const sockaddr* sender, socklen_t sender_len) override;
+		void receive_packet(BAN::ConstByteSpan, const sockaddr* sender, socklen_t sender_len) override;
 
-		virtual BAN::ErrorOr<void> bind_impl(const sockaddr* address, socklen_t address_len) override;
+		BAN::ErrorOr<void> connect_impl(const sockaddr*, socklen_t) override;
-		virtual BAN::ErrorOr<size_t> sendto_impl(BAN::ConstByteSpan message, const sockaddr* address, socklen_t address_len) override;
+		BAN::ErrorOr<void> bind_impl(const sockaddr* address, socklen_t address_len) override;
-		virtual BAN::ErrorOr<size_t> recvfrom_impl(BAN::ByteSpan buffer, sockaddr* address, socklen_t* address_len) override;
+		BAN::ErrorOr<size_t> recvmsg_impl(msghdr& message, int flags) override;
-		virtual BAN::ErrorOr<void> getpeername_impl(sockaddr*, socklen_t*) override { return BAN::Error::from_errno(ENOTCONN); }
+		BAN::ErrorOr<size_t> sendmsg_impl(const msghdr& message, int flags) override;
+		BAN::ErrorOr<void> getpeername_impl(sockaddr*, socklen_t*) override { return BAN::Error::from_errno(ENOTCONN); }
+		BAN::ErrorOr<void> getsockopt_impl(int, int, void*, socklen_t*) override;
+		BAN::ErrorOr<void> setsockopt_impl(int, int, const void*, socklen_t) override;
 
-		virtual bool can_read_impl() const override { return !m_packets.empty(); }
+		BAN::ErrorOr<long> ioctl_impl(int, void*) override;
-		virtual bool can_write_impl() const override { return true; }
+
-		virtual bool has_error_impl() const override { return false; }
+		bool can_read_impl() const override { return !m_packets.empty(); }
-		virtual bool has_hungup_impl() const override { return false; }
+		bool can_write_impl() const override { return true; }
+		bool has_error_impl() const override { return false; }
+		bool has_hungup_impl() const override { return false; }
 
 	private:
 		UDPSocket(NetworkLayer&, const Socket::Info&);
@@ -59,7 +64,10 @@ namespace Kernel
 		BAN::CircularQueue<PacketInfo, 32>	m_packets;
 		size_t								m_packet_total_size { 0 };
 		SpinLock							m_packet_lock;
-		ThreadBlocker							m_packet_thread_blocker;
+		ThreadBlocker						m_packet_thread_blocker;
+
+		sockaddr_storage					m_peer_address {};
+		socklen_t							m_peer_address_len { 0 };
 
 		friend class BAN::RefPtr<UDPSocket>;
 	};

kernel/include/kernel/Networking/UNIX/Socket.h

@@ -5,7 +5,8 @@
 #include <BAN/WeakPtr.h>
 #include <kernel/FS/Socket.h>
 #include <kernel/FS/TmpFS/Inode.h>
-#include <kernel/Lock/SpinLock.h>
+#include <kernel/FS/VirtualFileSystem.h>
+#include <kernel/OpenFileDescriptorSet.h>
 
 namespace Kernel
 {
@@ -15,6 +16,9 @@ namespace Kernel
 		BAN_NON_COPYABLE(UnixDomainSocket);
 		BAN_NON_MOVABLE(UnixDomainSocket);
 
+	public:
+		using FDWrapper = OpenFileDescriptorSet::FDWrapper;
+
 	public:
 		static BAN::ErrorOr<BAN::RefPtr<UnixDomainSocket>> create(Socket::Type, const Socket::Info&);
 		BAN::ErrorOr<void> make_socket_pair(UnixDomainSocket&);
@@ -24,9 +28,11 @@ namespace Kernel
 		virtual BAN::ErrorOr<void> connect_impl(const sockaddr*, socklen_t) override;
 		virtual BAN::ErrorOr<void> listen_impl(int) override;
 		virtual BAN::ErrorOr<void> bind_impl(const sockaddr*, socklen_t) override;
-		virtual BAN::ErrorOr<size_t> sendto_impl(BAN::ConstByteSpan, const sockaddr*, socklen_t) override;
+		virtual BAN::ErrorOr<size_t> recvmsg_impl(msghdr& message, int flags) override;
-		virtual BAN::ErrorOr<size_t> recvfrom_impl(BAN::ByteSpan, sockaddr*, socklen_t*) override;
+		virtual BAN::ErrorOr<size_t> sendmsg_impl(const msghdr& message, int flags) override;
 		virtual BAN::ErrorOr<void> getpeername_impl(sockaddr*, socklen_t*) override;
+		virtual BAN::ErrorOr<void> getsockopt_impl(int, int, void*, socklen_t*) override;
+		virtual BAN::ErrorOr<void> setsockopt_impl(int, int, const void*, socklen_t) override;
 
 		virtual bool can_read_impl() const override;
 		virtual bool can_write_impl() const override;
@@ -37,10 +43,8 @@ namespace Kernel
 		UnixDomainSocket(Socket::Type, const Socket::Info&);
 		~UnixDomainSocket();
 
-		BAN::ErrorOr<void> add_packet(BAN::ConstByteSpan);
+		bool is_bound() const { return !m_bound_file.canonical_path.empty(); }
-
+		bool is_bound_to_unused() const { return !m_bound_file.inode; }
-		bool is_bound() const { return !m_bound_path.empty(); }
-		bool is_bound_to_unused() const { return m_bound_path == "X"_sv; }
 
 		bool is_streaming() const;
 
@@ -53,7 +57,7 @@ namespace Kernel
 			BAN::WeakPtr<UnixDomainSocket>				connection;
 			BAN::Queue<BAN::RefPtr<UnixDomainSocket>>	pending_connections;
 			ThreadBlocker								pending_thread_blocker;
-			SpinLock									pending_lock;
+			Mutex										pending_lock;
 		};
 
 		struct ConnectionlessInfo
@@ -61,17 +65,31 @@ namespace Kernel
 			BAN::String peer_address;
 		};
 
+		struct PacketInfo
+		{
+			size_t size;
+			BAN::Vector<FDWrapper> fds;
+			BAN::Optional<struct ucred> ucred;
+			BAN::WeakPtr<UnixDomainSocket> sender;
+		};
+
+		BAN::ErrorOr<size_t> add_packet(const msghdr&, PacketInfo&&, bool dont_block);
+
 	private:
-		const Socket::Type	m_socket_type;
+		const Socket::Type		m_socket_type;
-		BAN::String			m_bound_path;
+		VirtualFileSystem::File	m_bound_file;
 
 		BAN::Variant<ConnectionInfo, ConnectionlessInfo> m_info;
 
-		BAN::CircularQueue<size_t, 128>	m_packet_sizes;
+		BAN::CircularQueue<PacketInfo, 512>	m_packet_infos;
-		size_t							m_packet_size_total { 0 };
+		size_t								m_packet_size_total { 0 };
-		BAN::UniqPtr<VirtualRange>		m_packet_buffer;
+		size_t								m_packet_buffer_tail { 0 };
-		SpinLock						m_packet_lock;
+		BAN::UniqPtr<VirtualRange>			m_packet_buffer;
-		ThreadBlocker					m_packet_thread_blocker;
+		mutable Mutex						m_packet_lock;
+		ThreadBlocker						m_packet_thread_blocker;
+
+		BAN::Atomic<size_t> m_sndbuf { 0 };
+		BAN::Atomic<size_t> m_bytes_sent { 0 };
 
 		friend class BAN::RefPtr<UnixDomainSocket>;
 	};

kernel/include/kernel/OpenFileDescriptorSet.h

@@ -33,7 +33,7 @@ namespace Kernel
 
 		BAN::ErrorOr<int> dup2(int, int);
 
-		BAN::ErrorOr<int> fcntl(int fd, int cmd, int extra);
+		BAN::ErrorOr<int> fcntl(int fd, int cmd, uintptr_t extra);
 
 		BAN::ErrorOr<off_t> seek(int fd, off_t offset, int whence);
 		BAN::ErrorOr<off_t> tell(int) const;
@@ -44,6 +44,10 @@ namespace Kernel
 		void close_all();
 		void close_cloexec();
 
+		bool is_cloexec(int fd);
+		void add_cloexec(int fd);
+		void remove_cloexec(int fd);
+
 		BAN::ErrorOr<void> flock(int fd, int op);
 
 		BAN::ErrorOr<size_t> read(int fd, BAN::ByteSpan);
@@ -51,8 +55,8 @@ namespace Kernel
 
 		BAN::ErrorOr<size_t> read_dir_entries(int fd, struct dirent* list, size_t list_len);
 
-		BAN::ErrorOr<size_t> recvfrom(int fd, BAN::ByteSpan buffer, sockaddr* address, socklen_t* address_len);
+		BAN::ErrorOr<size_t> recvmsg(int socket, msghdr& message, int flags);
-		BAN::ErrorOr<size_t> sendto(int fd, BAN::ConstByteSpan buffer, const sockaddr* address, socklen_t address_len);
+		BAN::ErrorOr<size_t> sendmsg(int socket, const msghdr& message, int flags);
 
 		BAN::ErrorOr<VirtualFileSystem::File> file_of(int) const;
 		BAN::ErrorOr<BAN::String> path_of(int) const;
@@ -74,8 +78,8 @@ namespace Kernel
 
 			struct flock_t
 			{
-				bool locked;
+				bool locked { false };
-				bool shared;
+				bool shared { false };
 				ThreadBlocker thread_blocker;
 				BAN::HashSet<pid_t> lockers;
 			};
@@ -84,36 +88,42 @@ namespace Kernel
 			friend class BAN::RefPtr<OpenFileDescription>;
 		};
 
-		struct OpenFile
-		{
-			OpenFile() = default;
-			OpenFile(BAN::RefPtr<OpenFileDescription> description, int descriptor_flags)
-				: description(BAN::move(description))
-				, descriptor_flags(descriptor_flags)
-			{ }
-
-			BAN::RefPtr<Inode> inode() const { ASSERT(description); return description->file.inode; }
-			BAN::StringView    path()  const { ASSERT(description); return description->file.canonical_path.sv(); }
-
-			int& status_flags() { ASSERT(description); return description->status_flags; }
-			const int& status_flags() const { ASSERT(description); return description->status_flags; }
-
-			off_t& offset() { ASSERT(description); return description->offset; }
-			const off_t& offset() const { ASSERT(description); return description->offset; }
-
-			BAN::RefPtr<OpenFileDescription> description;
-			int descriptor_flags { 0 };
-		};
-
 		BAN::ErrorOr<void> validate_fd(int) const;
 		BAN::ErrorOr<int> get_free_fd() const;
 		BAN::ErrorOr<void> get_free_fd_pair(int fds[2]) const;
 
+	public:
+		class FDWrapper
+		{
+		public:
+			FDWrapper(BAN::RefPtr<OpenFileDescription>);
+			FDWrapper(const FDWrapper& other) { *this = other; }
+			FDWrapper(FDWrapper&& other) { *this = BAN::move(other); }
+			~FDWrapper();
+
+			FDWrapper& operator=(const FDWrapper&);
+			FDWrapper& operator=(FDWrapper&&);
+
+			int fd() const { return m_fd; }
+
+			void clear();
+
+		private:
+			BAN::RefPtr<OpenFileDescription> m_description;
+			int m_fd { -1 };
+
+			friend class OpenFileDescriptorSet;
+		};
+
+		BAN::ErrorOr<FDWrapper> get_fd_wrapper(int fd);
+		size_t open_all_fd_wrappers(BAN::Span<FDWrapper> fd_wrappers);
+
 	private:
 		const Credentials& m_credentials;
 		mutable Mutex m_mutex;
 
-		BAN::Array<OpenFile, OPEN_MAX> m_open_files;
+		BAN::Array<BAN::RefPtr<OpenFileDescription>, OPEN_MAX> m_open_files;
+		BAN::Array<uint32_t, (OPEN_MAX + 31) / 32> m_cloexec_files {};
 	};
 
 }

kernel/include/kernel/PCI.h

@@ -187,7 +187,7 @@ namespace Kernel::PCI
 		void initialize_impl();
 
 	private:
-		static constexpr uint8_t m_msi_count = IRQ_SYSCALL - IRQ_MSI_BASE;
+		static constexpr uint8_t m_msi_count = IRQ_MSI_END - IRQ_MSI_BASE;
 		using PCIBus = BAN::Array<BAN::Array<Device, 8>, 32>;
 		BAN::Array<PCIBus, 256>  m_buses;
 		BAN::Array<paddr_t, 256> m_bus_pcie_paddr;

kernel/include/kernel/Panic.h

@@ -19,15 +19,26 @@ namespace Kernel
 		asm volatile("cli");
 
 		const bool had_debug_lock = Debug::s_debug_lock.current_processor_has_lock();
-		derrorln("Kernel panic at {}", location);
+
-		if (had_debug_lock)
+		bool first_panic = false;
-			derrorln("  while having debug lock...");
+
-		derrorln(message, BAN::forward<Args>(args)...);
-		if (!g_paniced)
 		{
-			g_paniced = true;
+			SpinLockGuard _(Debug::s_debug_lock);
-			Debug::dump_stack_trace();
+			derrorln("Kernel panic at {}", location);
+			if (had_debug_lock)
+				derrorln("  while having debug lock...");
+			derrorln(message, BAN::forward<Args>(args)...);
+			if (!g_paniced)
+			{
+				Debug::dump_stack_trace();
+				g_paniced = true;
+				first_panic = true;
+			}
 		}
+
+		if (first_panic)
+			Debug::dump_qr_code();
+
 		asm volatile("ud2");
 		__builtin_unreachable();
 	}

kernel/include/kernel/Process.h

@@ -9,6 +9,7 @@
 #include <kernel/ELF.h>
 #include <kernel/FS/Inode.h>
 #include <kernel/Lock/Mutex.h>
+#include <kernel/Lock/RWLock.h>
 #include <kernel/Memory/Heap.h>
 #include <kernel/Memory/MemoryRegion.h>
 #include <kernel/Memory/SharedMemoryObject.h>
@@ -76,6 +77,7 @@ namespace Kernel
 		BAN::ErrorOr<long> sys_getcwd(char* buffer, size_t size);
 		BAN::ErrorOr<long> sys_chdir(const char* path);
 		BAN::ErrorOr<long> sys_fchdir(int fildes);
+		BAN::ErrorOr<long> sys_chroot(const char* path);
 
 		BAN::ErrorOr<long> sys_setuid(uid_t);
 		BAN::ErrorOr<long> sys_setgid(gid_t);
@@ -95,6 +97,9 @@ namespace Kernel
 		BAN::ErrorOr<long> sys_getppid() const { return m_parent; }
 		BAN::ErrorOr<long> sys_getpid() const { return pid(); }
 
+		BAN::ErrorOr<long> sys_getgroups(gid_t groups[], size_t count);
+		BAN::ErrorOr<long> sys_setgroups(const gid_t groups[], size_t count);
+
 		BAN::ErrorOr<long> open_inode(VirtualFileSystem::File&&, int flags);
 
 		BAN::ErrorOr<void> create_file_or_dir(int fd, const char* path, mode_t mode) const;
@@ -105,9 +110,9 @@ namespace Kernel
 		BAN::ErrorOr<long> sys_access(const char* path, int amode);
 		BAN::ErrorOr<long> sys_create_dir(const char*, mode_t);
 		BAN::ErrorOr<long> sys_hardlinkat(int fd1, const char* path1, int fd2, const char* path2, int flag);
+		BAN::ErrorOr<long> sys_renameat(int oldfd, const char* old, int newfd, const char* _new);
 		BAN::ErrorOr<long> sys_unlinkat(int fd, const char* path, int flag);
 		BAN::ErrorOr<long> sys_readlinkat(int fd, const char* path, char* buffer, size_t bufsize);
-
 		BAN::ErrorOr<long> sys_symlinkat(const char* path1, int fd, const char* path2);
 
 		BAN::ErrorOr<long> sys_flock(int fd, int op);
@@ -130,8 +135,8 @@ namespace Kernel
 		BAN::ErrorOr<long> sys_bind(int socket, const sockaddr* address, socklen_t address_len);
 		BAN::ErrorOr<long> sys_connect(int socket, const sockaddr* address, socklen_t address_len);
 		BAN::ErrorOr<long> sys_listen(int socket, int backlog);
-		BAN::ErrorOr<long> sys_sendto(const sys_sendto_t*);
+		BAN::ErrorOr<long> sys_recvmsg(int socket, msghdr* message, int flags);
-		BAN::ErrorOr<long> sys_recvfrom(sys_recvfrom_t*);
+		BAN::ErrorOr<long> sys_sendmsg(int socket, const msghdr* message, int flags);
 
 		BAN::ErrorOr<long> sys_ioctl(int fildes, int request, void* arg);
 
@@ -142,10 +147,12 @@ namespace Kernel
 		BAN::ErrorOr<long> sys_epoll_ctl(int epfd, int op, int fd, struct epoll_event* event);
 		BAN::ErrorOr<long> sys_epoll_pwait2(int epfd, struct epoll_event* events, int maxevents, const struct timespec* timeout, const sigset_t* sigmask);
 
+		BAN::ErrorOr<long> sys_eventfd(unsigned int initval_hi, int flags);
+
 		BAN::ErrorOr<long> sys_pipe(int fildes[2]);
 		BAN::ErrorOr<long> sys_dup2(int fildes, int fildes2);
 
-		BAN::ErrorOr<long> sys_fcntl(int fildes, int cmd, int extra);
+		BAN::ErrorOr<long> sys_fcntl(int fildes, int cmd, uintptr_t extra);
 
 		BAN::ErrorOr<long> sys_seek(int fd, off_t offset, int whence);
 		BAN::ErrorOr<long> sys_tell(int fd);
@@ -161,35 +168,50 @@ namespace Kernel
 
 		BAN::ErrorOr<long> sys_sync(bool should_block);
 
+		BAN::ErrorOr<long> sys_get_nprocessor();
+
 		static BAN::ErrorOr<long> clean_poweroff(int command);
 		BAN::ErrorOr<long> sys_poweroff(int command);
 
 		BAN::ErrorOr<long> sys_readdir(int fd, struct dirent* list, size_t list_len);
 
+		BAN::ErrorOr<BAN::Vector<BAN::UniqPtr<MemoryRegion>>> split_memory_region(BAN::UniqPtr<MemoryRegion>&& region, vaddr_t base, size_t length);
 		BAN::ErrorOr<long> sys_mmap(const sys_mmap_t*);
 		BAN::ErrorOr<long> sys_munmap(void* addr, size_t len);
+		BAN::ErrorOr<long> sys_mprotect(void* addr, size_t len, int prot);
 		BAN::ErrorOr<long> sys_msync(void* addr, size_t len, int flags);
 
 		BAN::ErrorOr<long> sys_smo_create(size_t len, int prot);
 		BAN::ErrorOr<long> sys_smo_delete(SharedMemoryObjectManager::Key);
 		BAN::ErrorOr<long> sys_smo_map(SharedMemoryObjectManager::Key);
 
-		BAN::ErrorOr<long> sys_ttyname(int fildes, char* storage);
+		BAN::ErrorOr<long> sys_ttyname(int fildes, char* name, size_t namesize);
 		BAN::ErrorOr<long> sys_isatty(int fildes);
 		BAN::ErrorOr<long> sys_posix_openpt(int flags);
 		BAN::ErrorOr<long> sys_ptsname(int fildes, char* buffer, size_t buffer_len);
 
 		BAN::ErrorOr<long> sys_tty_ctrl(int fildes, int command, int flags);
 
-		static BAN::ErrorOr<void> kill(pid_t pid, int signal);
+		void set_stopped(bool stopped, int signal);
+		void wait_while_stopped();
+
+		static BAN::ErrorOr<void> kill(pid_t pid, int signal, const siginfo_t& = {});
 		BAN::ErrorOr<long> sys_kill(pid_t pid, int signal);
 		BAN::ErrorOr<long> sys_sigaction(int signal, const struct sigaction* act, struct sigaction* oact);
 		BAN::ErrorOr<long> sys_sigpending(sigset_t* set);
 		BAN::ErrorOr<long> sys_sigprocmask(int how, const sigset_t* set, sigset_t* oset);
+		BAN::ErrorOr<long> sys_sigsuspend(const sigset_t* set);
+		BAN::ErrorOr<long> sys_sigwait(const sigset_t* set, int* sig);
+		BAN::ErrorOr<long> sys_sigaltstack(const stack_t* ss, stack_t* oss);
 
+		BAN::ErrorOr<long> sys_futex(int op, const uint32_t* addr, uint32_t val, const timespec* abstime);
 		BAN::ErrorOr<long> sys_yield();
-		BAN::ErrorOr<long> sys_set_tls(void*);
+
-		BAN::ErrorOr<long> sys_get_tls();
+		BAN::ErrorOr<long> sys_set_fsbase(void*);
+		BAN::ErrorOr<long> sys_get_fsbase();
+		BAN::ErrorOr<long> sys_set_gsbase(void*);
+		BAN::ErrorOr<long> sys_get_gsbase();
+
 		BAN::ErrorOr<long> sys_pthread_create(const pthread_attr_t* attr, void (*entry)(void*), void* arg);
 		BAN::ErrorOr<long> sys_pthread_exit(void* value);
 		BAN::ErrorOr<long> sys_pthread_join(pthread_t thread, void** value);
@@ -199,8 +221,6 @@ namespace Kernel
 		BAN::ErrorOr<long> sys_tcgetpgrp(int fd);
 		BAN::ErrorOr<long> sys_tcsetpgrp(int fd, pid_t pgid);
 
-		BAN::ErrorOr<long> sys_termid(char*);
-
 		BAN::ErrorOr<long> sys_clock_gettime(clockid_t, timespec*);
 
 		BAN::ErrorOr<long> sys_load_keymap(const char* path);
@@ -209,24 +229,33 @@ namespace Kernel
 
 		static Process& current() { return Thread::current().process(); }
 
+		vaddr_t shared_page_vaddr() const { return m_shared_page_vaddr; }
+
 		PageTable& page_table() { return m_page_table ? *m_page_table : PageTable::kernel(); }
 
 		size_t proc_meminfo(off_t offset, BAN::ByteSpan) const;
 		size_t proc_cmdline(off_t offset, BAN::ByteSpan) const;
 		size_t proc_environ(off_t offset, BAN::ByteSpan) const;
+		BAN::ErrorOr<BAN::String> proc_cwd() const;
+		BAN::ErrorOr<BAN::String> proc_exe() const;
+
+		BAN::StringView executable() const { return m_executable; }
 
 		// Returns error if page could not be allocated
 		// Returns true if the page was allocated successfully
 		// Return false if access was page violation (segfault)
-		BAN::ErrorOr<bool> allocate_page_for_demand_paging(vaddr_t addr, bool wants_write);
+		BAN::ErrorOr<bool> allocate_page_for_demand_paging(vaddr_t addr, bool wants_write, bool wants_exec);
 
 		// FIXME: remove this API
 		BAN::ErrorOr<BAN::String> absolute_path_of(BAN::StringView) const;
 
+		OpenFileDescriptorSet& open_file_descriptor_set() { return m_open_file_descriptors; }
+
 		// ONLY CALLED BY TIMER INTERRUPT
 		static void update_alarm_queue();
 
 		const VirtualFileSystem::File& working_directory() const { return m_working_directory; }
+		const VirtualFileSystem::File& root_file() const { return m_root_file; }
 
 	private:
 		Process(const Credentials&, pid_t pid, pid_t parent, pid_t sid, pid_t pgrp);
@@ -245,13 +274,21 @@ namespace Kernel
 			BAN::StringView file_name;
 		};
 
+		// Adds new region to the sorted array of mapped regions
+		// You must hold writer end of m_mapped_region_lock when calling this.
+		BAN::ErrorOr<void> add_mapped_region(BAN::UniqPtr<MemoryRegion>&&);
+		// If address is contained by a region, returns the index of that.
+		// Otherwise returns the address of the first region after this address.
+		// You must hold reader end of m_mapped_region_lock when calling this.
+		size_t find_mapped_region(vaddr_t) const;
+
 		BAN::ErrorOr<VirtualFileSystem::File> find_file(int fd, const char* path, int flags) const;
 		BAN::ErrorOr<FileParent> find_parent_file(int fd, const char* path, int flags) const;
 		BAN::ErrorOr<VirtualFileSystem::File> find_relative_parent(int fd, const char* path) const;
 
-		BAN::ErrorOr<void> validate_string_access(const char*);
+		BAN::ErrorOr<void> read_from_user(const void* user_addr, void* out, size_t size);
-		BAN::ErrorOr<void> validate_pointer_access_check(const void*, size_t, bool needs_write);
+		BAN::ErrorOr<void> read_string_from_user(const char* user_addr, char* out, size_t max_size);
-		BAN::ErrorOr<void> validate_pointer_access(const void*, size_t, bool needs_write);
+		BAN::ErrorOr<void> write_to_user(void* user_addr, const void* in, size_t size);
 		BAN::ErrorOr<MemoryRegion*> validate_and_pin_pointer_access(const void*, size_t, bool needs_write);
 
 		uint64_t signal_pending_mask() const
@@ -260,7 +297,7 @@ namespace Kernel
 			return m_signal_pending_mask;
 		}
 
-		void add_pending_signal(uint8_t signal)
+		void add_pending_signal(uint8_t signal, const siginfo_t& info)
 		{
 			ASSERT(signal >= _SIGMIN);
 			ASSERT(signal <= _SIGMAX);
@@ -272,6 +309,7 @@ namespace Kernel
 			if (handler == SIG_DFL && (signal == SIGCHLD || signal == SIGURG))
 				return;
 			m_signal_pending_mask |= 1ull << signal;
+			m_signal_infos[signal] = info;
 		}
 
 		void remove_pending_signal(uint8_t signal)
@@ -284,18 +322,18 @@ namespace Kernel
 		}
 
 	private:
-		struct ChildExitStatus
+		struct ChildWaitStatus
 		{
-			pid_t    pid       { 0 };
+			pid_t pid  { 0 };
-			pid_t    pgrp      { 0 };
+			pid_t pgrp { 0 };
-			int      exit_code { 0 };
+			BAN::Optional<int> status;
-			bool     exited    { false };
 		};
 
 		Credentials m_credentials;
 
 		OpenFileDescriptorSet m_open_file_descriptors;
 
+		mutable RWLock m_memory_region_lock;
 		BAN::Vector<BAN::UniqPtr<MemoryRegion>> m_mapped_regions;
 
 		pid_t m_sid;
@@ -305,7 +343,27 @@ namespace Kernel
 
 		mutable Mutex m_process_lock;
 
+		BAN::Atomic<bool> m_stopped { false };
+		ThreadBlocker m_stop_blocker;
+
 		VirtualFileSystem::File m_working_directory;
+		VirtualFileSystem::File m_root_file;
+
+		vaddr_t m_shared_page_vaddr { 0 };
+
+		struct futex_t
+		{
+			Mutex mutex;
+			ThreadBlocker blocker;
+			uint32_t waiters { 0 };
+			uint32_t to_wakeup { 0 };
+		};
+
+		static BAN::HashMap<paddr_t, BAN::UniqPtr<futex_t>> s_futexes;
+		static Mutex s_futex_lock;
+
+		BAN::HashMap<paddr_t, BAN::UniqPtr<futex_t>> m_futexes;
+		Mutex m_futex_lock;
 
 		BAN::Vector<Thread*> m_threads;
 
@@ -321,20 +379,26 @@ namespace Kernel
 		uint64_t m_alarm_wake_time_ns { 0 };
 
 		mutable SpinLock m_signal_lock;
+		siginfo_t        m_signal_infos[_SIGMAX + 1] { };
 		struct sigaction m_signal_handlers[_SIGMAX + 1] { };
 		uint64_t m_signal_pending_mask { 0 };
 
 		BAN::Vector<BAN::String> m_cmdline;
 		BAN::Vector<BAN::String> m_environ;
+		BAN::String m_executable;
 
-		BAN::Vector<ChildExitStatus> m_child_exit_statuses;
+		BAN::Vector<ChildWaitStatus> m_child_wait_statuses;
-		ThreadBlocker m_child_exit_blocker;
+		SpinLock m_child_wait_lock;
+		ThreadBlocker m_child_wait_blocker;
+
+		BAN::Atomic<bool> m_is_exiting { false };
 
 		bool m_has_called_exec { false };
 
 		BAN::UniqPtr<PageTable> m_page_table;
 		BAN::RefPtr<TTY> m_controlling_terminal;
 
+		friend class OpenFileDescriptorSet;
 		friend class Thread;
 	};
 

kernel/include/kernel/Processor.h

@@ -3,10 +3,12 @@
 #include <BAN/Atomic.h>
 #include <BAN/ForwardList.h>
 
+#include <kernel/API/SharedPage.h>
 #include <kernel/Arch.h>
 #include <kernel/GDT.h>
 #include <kernel/IDT.h>
 #include <kernel/InterruptStack.h>
+#include <kernel/Memory/Types.h>
 #include <kernel/ProcessorID.h>
 #include <kernel/Scheduler.h>
 
@@ -33,6 +35,7 @@ namespace Kernel
 				FlushTLB,
 				NewThread,
 				UnblockThread,
+				UpdateTSC,
 				StackTrace,
 			};
 			SMPMessage* next { nullptr };
@@ -43,6 +46,7 @@ namespace Kernel
 				{
 					uintptr_t vaddr;
 					size_t page_count;
+					void* page_table;
 				} flush_tlb;
 				SchedulerQueue::Node* new_thread;
 				SchedulerQueue::Node* unblock_thread;
@@ -55,17 +59,19 @@ namespace Kernel
 		static Processor& initialize();
 
 		static ProcessorID current_id() { return read_gs_sized<ProcessorID>(offsetof(Processor, m_id)); }
+		static uint8_t current_index()  { return read_gs_sized<uint8_t>(offsetof(Processor, m_index)); }
 		static ProcessorID id_from_index(size_t index);
 
-		static uint8_t count()       { return s_processor_count; }
+		static uint8_t count()        { return s_processor_count; }
-		static bool is_smp_enabled() { return s_is_smp_enabled; }
+		static bool is_smp_enabled()  { return s_is_smp_enabled; }
+		static void set_smp_enabled() { s_is_smp_enabled = true; }
 		static void wait_until_processors_ready();
 
 		static void toggle_should_print_cpu_load() { s_should_print_cpu_load = !s_should_print_cpu_load; }
 		static bool get_should_print_cpu_load() { return s_should_print_cpu_load; }
 
-		static ProcessorID bsb_id() { return s_bsb_id; }
+		static ProcessorID bsp_id() { return s_bsp_id; }
-		static bool current_is_bsb() { return current_id() == bsb_id(); }
+		static bool current_is_bsp() { return current_id() == bsp_id(); }
 
 		static void set_interrupt_state(InterruptState state)
 		{
@@ -77,8 +83,11 @@ namespace Kernel
 
 		static InterruptState get_interrupt_state()
 		{
-			uintptr_t flags;
+#if ARCH(x86_64)
-			asm volatile("pushf; pop %0" : "=rm"(flags));
+			const auto flags = __builtin_ia32_readeflags_u64();
+#elif ARCH(i686)
+			const auto flags = __builtin_ia32_readeflags_u32();
+#endif
 			if (flags & (1 << 9))
 				return InterruptState::Enabled;
 			return InterruptState::Disabled;
@@ -97,6 +106,8 @@ namespace Kernel
 		uintptr_t stack_bottom() const	{ return reinterpret_cast<uintptr_t>(m_stack); }
 		uintptr_t stack_top() const		{ return stack_bottom() + s_stack_size; }
 
+		static void set_thread_syscall_stack(vaddr_t vaddr) { write_gs_sized<vaddr_t>(offsetof(Processor, m_thread_syscall_stack), vaddr); }
+
 		static GDT& gdt() { return *read_gs_sized<GDT*>(offsetof(Processor, m_gdt)); }
 		static IDT& idt() { return *read_gs_sized<IDT*>(offsetof(Processor, m_idt)); }
 
@@ -106,13 +117,40 @@ namespace Kernel
 		static void yield();
 		static Scheduler& scheduler() { return *read_gs_sized<Scheduler*>(offsetof(Processor, m_scheduler)); }
 
+		static void initialize_tsc(uint8_t shift, uint64_t mult, uint64_t realtime_seconds);
+		static void update_tsc();
+		static uint64_t ns_since_boot_tsc();
+
+		static Thread* get_current_sse_thread() { return read_gs_sized<Thread*>(offsetof(Processor, m_sse_thread)); };
+		static void set_current_sse_thread(Thread* thread) { write_gs_sized<Thread*>(offsetof(Processor, m_sse_thread), thread); };
+
+		static paddr_t shared_page_paddr() { return s_shared_page_paddr; }
+		static volatile API::SharedPage& shared_page() { return *reinterpret_cast<API::SharedPage*>(s_shared_page_vaddr); }
+
 		static void handle_ipi();
 
 		static void handle_smp_messages();
 		static void send_smp_message(ProcessorID, const SMPMessage&, bool send_ipi = true);
 		static void broadcast_smp_message(const SMPMessage&);
 
-		static void load_tls();
+		static void load_segments();
+		static void load_fsbase();
+		static void load_gsbase();
+
+		static void disable_sse()
+		{
+			uintptr_t dummy;
+#if ARCH(x86_64)
+			asm volatile("movq %%cr0, %0; orq $0x08, %0; movq %0, %%cr0" : "=r"(dummy));
+#elif ARCH(i686)
+			asm volatile("movl %%cr0, %0; orl $0x08, %0; movl %0, %%cr0" : "=r"(dummy));
+#endif
+		}
+
+		static void enable_sse()
+		{
+			asm volatile("clts");
+		}
 
 	private:
 		Processor() = default;
@@ -120,6 +158,16 @@ namespace Kernel
 
 		static ProcessorID read_processor_id();
 
+		static void initialize_smp();
+		static void initialize_shared_page();
+
+		static void dummy()
+		{
+#if ARCH(x86_64)
+			static_assert(offsetof(Processor, m_thread_syscall_stack) == 8, "This is hardcoded in Syscall.S");
+#endif
+		}
+
 		template<typename T>
 		static T read_gs_sized(uintptr_t offset) requires(sizeof(T) <= 8)
 		{
@@ -153,12 +201,19 @@ namespace Kernel
 		}
 
 	private:
-		static ProcessorID s_bsb_id;
+		static ProcessorID s_bsp_id;
 		static BAN::Atomic<uint8_t> s_processor_count;
 		static BAN::Atomic<bool>    s_is_smp_enabled;
 		static BAN::Atomic<bool>    s_should_print_cpu_load;
+		static paddr_t              s_shared_page_paddr;
+		static vaddr_t              s_shared_page_vaddr;
 
-		ProcessorID m_id { PROCESSOR_NONE };
+		ProcessorID m_id { 0 };
+		uint8_t m_index { 0 };
+
+		vaddr_t m_thread_syscall_stack;
+
+		Thread* m_sse_thread { nullptr };
 
 		static constexpr size_t s_stack_size { 4096 };
 		void* m_stack { nullptr };
@@ -173,12 +228,8 @@ namespace Kernel
 		uint64_t m_last_update_ns { 0 };
 		uint64_t m_next_update_ns { 0 };
 
-		BAN::Atomic<bool> m_smp_pending_lock { false };
+		BAN::Atomic<SMPMessage*> m_smp_pending { nullptr };
-		SMPMessage* m_smp_pending { nullptr };
+		BAN::Atomic<SMPMessage*> m_smp_free    { nullptr };
-
-		BAN::Atomic<bool> m_smp_free_lock { false };
-		SMPMessage* m_smp_free    { nullptr };
-
 		SMPMessage* m_smp_message_storage { nullptr };
 
 		void* m_current_page_table { nullptr };